Issuu on Google+

M E E T I N G OF THE

Minds

U n d e r g r a d u a t e Research Symposium


Meeting of the minds is an annual symposium at Carnegie Mellon University that gives students an opportunity to present their research and project work to a wide audience of faculty, fellow students, family members, industry representatives and the larger community. Students use posters, videos and other visual aides to present their work in a manner that can be easily understood by both experts and non experts.

Through this experience, students learn how to brindege the gap between conducting research and presenting it to a wider audience. A review committee consisting of industry experts and faculty members from other universities will review the presentations and choose the best projects and posters. Awards and certificates are presented to the winners.


Table of Contents

POSTER # TITLE

PAGE

Biological Science Posters

Q1 Adherence of Pathogenic Fungi From the Qatari Clinical Setting

1

Q2 Biofilm Formation of Pathogenic Fungi Isolated at Hamad Hospital

3

Q3 Characterizing the Morphology of Pathogenic Fungi Isoloated in the Qatari Hospital Setting

5

Q4 Detection of Genetically Modified Organisms in Food Products

7

Business Administration Posters

Q5 Investigation of Legal & Regulatory Obstacles to Becoming an Entrepreneur in Qatar

Q6 Home-Based Business: A Growing Phenomenon

9 11

Computer Science Posters

Q7 An Authorization Model for the Web Programming Language Qwel

13

Q8 Arabic Accented Facial Expressions for a 3D Agent

15

Q9 CMUQ Official Android App

17

Q10 Interaction Analysis of Multi-lingual Robot Receptionist

19

Q11 I want my Mommy

21

Q12 Semiotic Circles: An Inclusive Methodology for Human Computer Interaction Design

23

Q13 Multi-Robot Coordination

25

Q14 Towards Computational Offloading in Mobile Clouds

27

Q15 PyExoPlanets: A Computed Application for Detecting Exoplanets’ Transits

on Stars’ Light Curves Q16 Unsupervised Word Segmentation and Statistical Machine Translation

29 31

Information Systems Posters

Q17 What Affects Students’ Acceptance & Use of Technology?

33

Q18 Software Development Project Electronic Resolution

35

Q19 Privacy and e-commerce in the Arab Culture

37

Q20 The Effect of type of Display on Conjoint Studies

39

General Education Posters

Q21 Arab Spring Newspaper Coverage

41

Q22 Let the Grades Flow!

43

Post-Graduate Posters

QG1 Decentralized Execution of Multiset Rewriting Rules of Ensembles

45

QG2 ClusterLoc: Exploiting Short Range Wireless Technologies for Energy Efficient Localization

47

QG3 Characterization of MapReduce Applications on Private and Public Cloud Platforms

49

QG4 The QALB Project: Building Resources and Systems for the Automatic Correction of Arabic Text 51

QG5 Type-Based Productivity of Stream Definitions

53


Tuesday, April 30, 2013, 4:00 pm - 6:00 pm Carnegie Mellon University, Education City


Adherence of Pathogenic Fungi from the Qatari Clinical Setting Author Mei ElGindi (BS 2013)

Faculty Advisor Jonathan Finkel, Ph.D.

Category Biological Sciences

Abstract Fungal infections are becoming a major source of secondary infections in the clinical setting, for diabetics and for immune compromised individuals. The majority of infections are thought to be the consequence of the fungi’s ability to form a microcommunity of cells acting in concert as a biofilm, which can form on the surface of medically implanted devices. The biofilm is surrounded by an extracellular matrix resulting in resistance to antifungal drugs. Each fungus presents itself in a different form inside the host. From the morphology of the species, to the symptoms that they cause, every fungus is distinct. Therefore, it is critical to study these fungi and determine the characteristics they present. This will help better identify the specific fungus infecting the patient and will allow for a faster and more precise treatment. The first step of biofilm formation is adherence of yeast cells to a surface. To better understand the differences between various strains of pathogenic fungi, we examined the adherence of patient strains obtained from Hamad Medical Corporation. These fungi include Candida spp., Trichosporon spp., and Geotrichum spp. The adherence of the different species was determined using an established adherence assay that measures the specific adherence within a single species. This difference in adherence was also used to detect any differences in adherence due to the site of infection.

1


Adherence of Pathogenic Fungi from the Qatari Clinical Setting Mei ElGindi and Jonathan S. Finkel Ph.D. Department of Biological Sciences, Carnegie Mellon University Qatar

Abstract

Adherence of Different Wild Type Clinical Strains

Fungal infections are becoming a major source of secondary infections in the clinical setting, for diabetics and for immune compromised individuals. The majority of infections are thought to be the consequence of the fungi’s ability to form a microcommunity of cells acting in concert as a biofilms, which can form on the surface of medically implanted devices. The biofilm is surrounded by an extracellular matrix resulting in resistance to antifungal drugs. Each fungus presents itself in a different form inside the host. From the morphology of the species, to the symptoms that they cause, every fungus is distinct. Therefore, it is critical to study these fungi and determine the characteristics they present. This will help better identify the specific fungus infecting the patient and will allow for a faster and more precise treatment. The first step of biofilm formation is adherence of yeast cells to a surface. To better understand the differences between various strains of pathogenic fungi, we examined the adherence of patient strains obtained from Hamad Medical Corporation. These fungi include Candida spp., Trichosporon spp., and Geotrichum spp. The adherence of the different species was determined using an established adherence assay that measures the specific adherence within a single species. This difference in adherence was also used to detect any differences in adherence due to the site of infection.

1.2  

CA = Candida albicans CD = Candida dubliniensis CT = Candida tropicalis TA = Trichosporon asahii TD = Trichosporon dohaense CP = Candida parapsilosis

1  

0.8  

0.6  

0.4  

0.2  

0  

CA  

CD  

CT  

TA  

TD  

CP  

Adherence of Different Candida albicans Patient Strains 2.5  

CAS = Candida albicans wild type Strain CAp = Candida albicans Patient Strain 2  

1.5  

1  

Steps of Biofilm Development

0.5  

In the first step, yeast form cells adhere to the substrate. Following this, the cells divide and multiply. In the third step, the biofilm enlarges with an increased number of yeast cells, hyphal cells, and extracellular matrix that envelops the biofilm. Finally, the yeast cells are dispersed to further colonize the surrounding environment or host.

0  

  CAS  

CAp1  

CAp2  

CAp3  

CAp4  

CAp5  

CAp6  

CAp7  

CAp8  

CAp9  

CAp10  

Adherence of different Candida dubliniensis Patient Strains

Fluxion 200 Bioflux Machine

1.6  

CDS = Candida dubliniensis wild type Strain CDp = Candida dubliniensis Patient Strain  

1.4  

² Contains an embedded micron-scale fluidic channel ² Allows for control of the flow across the channel ² Adhesion can be visualized and studied

1.2  

1  

0.8  

0.6  

0.4  

http://www.fluxionbio.com/ 0.2  

0  

After 30 Minutes

Methods

² Overnight  cultures  grown  in  YPD  (rich  growth  media)  are  diluted  to  OD600  =  0.2 ² The cell are added to the well at a flow rate of 3 dyn/cm3 ² Cells are allowed to adhere under flow for 30 minutes ² Four images are taken in each well ² These images are analyzed and the number of adhered cells are counted and compared to a wild type laboratory strain

  CDS  

CDp1  

CDp2  

CDp3  

CDp4  

CDp5  

CDp6  

CDp7  

CDp8  

CDp9  

CDp10  

CDp11  

CDp12  

CDp13  

Conclusions ² Adherence corresponds to approximate occurrence in the clinical settings ² Qatar patient strains were isolated from the lungs of Cystic Fibrosis patients, with C. dubliniensis being the most prevalently identified fungi ² The Cystic Fibrosis fungal species showed varying degrees of adherence, indicating that this ability may not be required for fungal lung pathogenicity

Acknowledgments We would like to thank the members of the lab for their assistance – Dr. Annette Vincent, Dr. Kenneth Hovis, Ridin Balakrishnan, Raji Katibe. This project was generously funded by Carnegie Mellon University Qatar and Qatar Foundation


Biofilm Formation of Pathogenic Fungi Isolated at Hamad Hospital

Author Raji Katibe (BS 2013)

Faculty Advisor Jonathan Finkel, Ph.D.

Category Biological Sciences

Abstract: Infection by pathogenic fungi has a 30% mortality rate in the United States alone. Furthermore, fungal infections cost the U.S. healthcare system around 2.6 billion dollars annually. In Qatar, continuous ambulatory peritoneal dialysis fungal infections occur in up to 16% of all patients. The formation of biofilms on catheter, artificial joints and artificial heart valves is a cause of systemic infection in the patients. Moreover, the result of the biofilm formation may be fatal infections. Many pathogenic yeast have the ability to form biofilm, where they form colonies and attach to organic and non-organic surfaces. The research described below focused on biofilm formation of three pathogenic fungi species: Trichosporon spp., Candida spp., and Geotrichum spp. isolated from patients at Hamad Medical Center (HMC). Biofilm formation usually occurs in four steps. First, in the adhesion step, yeast cells adhere to a surface, then, in the initiation step, the yeast cells start proliferating and switch morphology to elongated filaments called hyphae. In the third step, yeast and hyphal cells proliferate and extracellular matrix is formed. Finally, in the dispersal step, yeast cells are dispersed from the biofilm to further infect other areas of the body. Biofilm formation was assayed by allowing fungi to form biofilm in conditions similar to the human body. We began by isolating the species that have the ability to form biofilms from Hamad Hospital. Then, we studied the biofilms by comparing the biofilms formed by different species. We also compared the biofilm formed by the same species in different locations of infection.

3


Biofilm Formation of Pathogenic Fungi isolated at Hamad Hospital Raji  Ka'be,  and  Jonathan  S.  Finkel  Ph.D.   Department  of  Biological  Sciences  Carnegie  Mellon  University,  Qatar  

Result  

Abstract     Infec'on   by   pathogenic   fungi   has   a   30%   mortality   rate   in   the   United   States   alone.   Furthermore,   fungal   infec'ons   cost   the   US   healthcare   around   2.6   billion   dollars   annually.  In  Qatar,  pa'ents  suffering  from  con'nuous  ambulatory  peritoneal  dialysis   fungal   infec'ons   occurs   in   up   to   16%   of   all   pa'ents.   The   forma'on   of   biofilms   on   catheter,   ar'ficial   joints   and   ar'ficial   heart   valves   is   a   cause   of   systemic   infec'on   in   the   pa'ents.   Moreover,   the   result   of   the   biofilm   forma'on   may   be   fatal   infec'ons.   Many  pathogenic  yeast  have  the  ability  to  form  biofilms  that  they  use  to  form  colonies   and  aIach  to  organic  and  non-­‐organic  surfaces.  The  research  described  below  focused   on   biofilm   forma'on   of   three   pathogenic   fungi   species:   Trichosporon   spp.,   Candida   spp.,   and   Geotrichum   spp.     isolated   from   pa'ents   at   Hamad   Medical   Center   (HMC).   Biofilm   forma'on   usually   occurs   in   four   steps.   First   in   the   adhesion   step,   yeast   cells   adhere   to   a   surface.   Next   in   the   ini'a'on   step,   the   yeast   starts   prolifera'ng   and   switch  morphology  to  elongated  filaments  called  hyphae.  In  the  third  step,  yeast  and   hyphal   cells   proliferate   and   extracellular   matrix   is   formed.   Finally,   in   the   dispersal   step,   yeast   cells   are   dispersed   from   the   biofilm   to   further   infect   other   areas   of   the   host.   Biofilm   forma'on   was   assayed   by   allowing   fungi   to   form   biofilms   in   condi'ons   similar   to   the   human   body.   We   began   by   observing   for   biofilm   forma'on   in   fungal     species   collected   from   Hamad   Medical   Center.   Then,   we   studied   the   biofilms   by   comparing   the   biofilms   formed   by   different   species.   We   also   compared   the   biofilm   formed  by  the  same  species  in  different  loca'ons  of  infec'on.    

Nega've  Control  

Candida  albicans  

Candida  dubliniensis  

Trichosporon  mucoides  

Trichosporon  asahii    

Trichosporon  dohaense  

Candida  glabrata  

Candida  parapsilosis  

Candida  krusei  

The  Steps  of  Biofilm  Development  

Extracellular Matrix

Candida albicans

Adherence

Candida  tropicalis  

hyphae

Yeast form cells

Initiation

Maturation

Dispersal

Trichosporon spp. arthroconidia

Initiation

Adherence

Maturation

Dispersal Dispersal

Geotrichum spp.

Initiation

Adherence

•  •  Maturation

Dispersal

• 

Dr  Jonathan  Finkel  “Adherence  and  biofilm  forma'on  of  pathogenic  fungi  from  the  Qatari  clinical  seZng”  

• 

Methods   The   different   fungal   species   were   incubated   in   wells   containing   silicon   squares.   The   samples   were   then   incubated   at   37oC   to   allow   the   cells   to   adhere   to   the   silicon   squares.   A^er   90   minutes   the   squares   were   washed   with   1X   PBS   and   incubated   in   RPMI   for   48   hours   to   allow   the   biofilm   forma'on.   RPMI   is   a   media   that   allows   for   biofilm   forma'on   in   many   fungal   species.   The   biofilms   formed   on   the   squares   are   analyzed  and  compared  to  the  different  fungal  species.  The  weight,  height  and  shape   of  the  biofilms  will  be  studied  and  compared  to  the  different  species.  The  biofilms  will   be   studied   under   a   scanning   electron   microscope   and   a   confocal   microscope   so   that   the  structure  and  architecture  of  the  biofilm  can  be  observed.    

1. Add RPMI

2. Add Fungal Cells Incubate 37oC 90 minutes

Conclusions  

Only  6  out  of  our  9  fungal  samples  were  able  to  form  biofilms   C.   glabrata,   C.   parapsilosis   and   C.   tropicalis   failed   to   form   biofilms,   contrary  to  reports  in  the  literature   RPMI  did  not  create  the  biofilms  we  were  hoping  to  see,  the  biofilms   were  grainy  and  not  uniform  in  some  of  the  samples.  This  will  require   that   we   aIempt   to   observe   biofilms   in   other   condi'ons   such   as   M199,  YPD,  and  Spider  media   Candida  albicans  forms  the  largest  biofilm  of  the  species  examined,   as  expected  

Future  DirecAons  

•  The   weight   and   height   of   the   biofilms   formed   will   be   examined     using  confocal  microscopy     •  The   structure   and   architecture   of   the   biofilm   will   be   studied   under   a   scanning   electron   microscope,   which   will   provided   detailed  structural  images   •  Pa'ent   isolates   will   be   compared   to   the   wild   type   laboratory   strains  for  differences  within  species   •  Pa'ent   isolates   from   Qatar   will   also   be   used   to   determine   any   differences  that  may  occur  within  species  obtained  from  different   sites  of  infec'on  

Acknowledgements   1.  Wash  PBS     2.  RPMI  48  hrs    

I  thank  Dr.  Jonathan  S.  Finkel,  Dr.  AnneIe  Vincent,  Dr.  Kenneth  Hovis,  Maria   Navaro,  Ridin  Balakrishnan,  Mei  ElGindi  for  their  assistance.  I  would  also  like   to   thank   Carnegie   Mellon   University   Qatar   and   Qatar   Founda'on   for   their   generous  funding.    


Characterizing the Morphology of Pathogenic Fungi Isolated in the Qatari Hospital Setting Author Ridin Balakrishnan (BS 2013)

Faculty Advisor Jonathan Finkel, Ph.D.

Category Biological Sciences

Abstract: Fungi are opportunistic pathogens that have the ability to cause various superficial and systemic infections. Their ability to change morphology is strategic for virulence and biofilm formation. The fungal species Candida albicans is the primary cause of invasive fungal infections from yeasts. Nevertheless, a growing number of new infections from non-albicans Candida species are increasingly being recognized as a major source of infection, especially in hospitals. Our study focuses on Trichosporon spp., Candida spp. and Geotrichum spp. isolated from patients at the Qatari Hospital, Hamad Medical Center. Fungal infections from these species have been on the rise since the introduction of the C. albicans-specific antifungal caspofungin. These infections are much more prevalent in diabetics and immunocompromised patients. In addition to increasing mortality rates, these infections also result in increased healthcare costs and reduce productivity. Moreover, fungal infections can be caused by fungi that naturally occur in the soil which can easily get airborne during construction projects. Given the high prevalence of diabetics in Qatar and the current construction boom, the study of these opportunistic fungi in the context of Qatar’s medical environment is of the utmost importance. The goal of this research endeavor was to determine and characterize the morphology of various pathogenic fungi isolated in the clinical setting. Specifically, we studied three areas of differences in morphology: (1) the different species of fungi with regard to pathogenicity, (2) comparison due to the different sites of infection, (3) pathogenic fungi from the same species, but different patients at the Hamad Medical Center. Differences were compared with the stock ‘wild-type’ strains and between the different samples using phase contrast microscopy.

5


Characterizing  the  Morphology  of  Pathogenic  Fungi  Isolated   in  the  Qatari  Hospital  SeAng   Ridin  Balakrishnan  

 Jonathan  Finkel,  Ph.D.  

 Department  of  Biological  Sciences,    Carnegie  Mellon  University-­‐Qatar   Fungal  Species  

Abstract   Fungi  are  opportunis.c  pathogens  that  have  the  ability  to  cause  various  superficial  and  systemic  infec.ons.   Their  ability  to  change  morphology  is  strategic  for  virulence  and  biofilm  forma.on.  The  fungal  species  Candida   albicans   is   the   primary   cause   of   invasive   fungal   infec.ons   from   yeasts.   Nevertheless,   a   growing   number   of   new   infec.ons   from   non-­‐albicans   Candida   species   are   increasingly   being   recognized   as   a   major   source   of   infec.on,  especially  in  hospitals.  Our  study  focuses  on  Trichosporon  spp.,  Candida  spp.  and  Geotrichum  spp.   isolated   from   pa.ents   at   the   Qatari   Hospital,   Hamad   Medical   Center.   Fungal   infec.ons   from   these   species   have   been   on   the   rise   since   the   introduc.on   of   the   C.   albicans   specific   an.fungal   caspofungin.   These   infec.ons  are  much  more  prevalent  in  diabe.cs  and  immunocompromised  pa.ents.  In  addi.on  to  increasing   mortality  rates,  these  infec.ons  also  result  in  increased  healthcare  costs  and  reduce  produc.vity.  Moreover,   fungal  infec.ons  can  be  caused  by  fungi  that  naturally  occur  in  the  soil  which  can  easily  get  airborne  during   construc.on  projects.  Given  the  high  prevalence  of  diabe.cs  in  Qatar  and  the  current  construc.on  boom,  the   study  of  these  opportunis.c  fungi  in  the  context  of  Qatar’s  medical  environment  is  of  the  utmost  importance.   The  goal  of  this  research  endeavor  was  to  determine  and  characterize  the  morphology  of  various  pathogenic   fungi  isolated  in  the  clinical  seJng.  Specifically,  we  studied  differences  in  morphology  based  on  the  various   species   of   fungi   with   regard   to   pathogenicity.     We   used   pa.ent   strains   commonly   employed   in   the   laboratory   seJng   as   wild   type   strains.   The   comparisons   were   made   to   iden.fy   any   possible   morphological   differences   between   strains   and   as   the   basis   for   possible   morphological   architectural   variances   within   biofilms   formed   by   the  respec.ve  strains.    

    Background:      

Different  morphologies  of  Yeast  

                             Yeast  Form                                                        Pseudohyphal                                                  Hyphal                                                                                                                                                                      Filamentous    

                       Arthroconidial  

•  The  ability  to  filament  is  essen.al  for  infec.on  by  pathogenic  fungi.   •  The  yeast  form  cells  are  thought  to  be  the  commensal  form  and  used  in  biofilms  to  further   propagate  in  a  host  before  becoming  virulent.   •  The  fungi  contain  mechanisms  to  switch  between  these  different  forms.      

 

 

Methods:     3hrs  incuba.on       +  shaking            

 

   

10  ul  

-­‐  Various  Yeast    strains  grown    at    24oC/37oC  for  3-­‐4  hours.   -­‐  YPD  used  to  grow    at  24oC.   -­‐  RPMI/M199  used  to  grow  at  37oC.   -­‐      Thorough  shaking.   -­‐  10  ul  observed  using  phase  contrast  microscopy.   -­‐  Various  morphologies  analyzed.  

     

Phase  Contrast   Microscopy  

32oC  

37oC  

  C.  albicans       C.  dubliniensis       C.  krusei       C.  tropicalis         C.  parapsilosis       C.  glabrata       T.  mucoides       T.  asahii       T.  dohaense       G.  candidum       G.  capitatum         Conclusions         Majority   of   the   fungi   appear   to   have   at   least   two   forms   of   morphologies   (yeast  form  &  filamentous)  as  expected.    The  last  three    fungi  all  had  liQle    to   no   yeast   form   cells.   This   may   be   one   of   the   reasons   that   they   are   observed   less  frequently  in  hospital  seJngs.  The  results  of  this  experiment     have   shown   us     what   to   expect   when   we   look   at   different   pathogenic     in  the  future.  Lastly,  it  has  also  given  us  an  idea  of  what  kind  of  cells   forms   biofilms   will  be  composed  of,  in  the  clinical  seJng.       Acknowledgments       We  thank  the  members  of  the  lab  for  their  assistance  –Ms.  Maria  Navaro,  Dr.  AnneQe  Vincent,  Dr.  Kenneth  Hovis,  Mei  ElGindi,  Raji   Ka.be.  This  project  was  funded  by  Carnegie  Mellon  University  Qatar  and  Qatar  Founda.on  and  we  thank  them  for  their  generosity.  

         


Detection of Genetically Modified Organisms in Food Products Authors Hiba Al-Ashtal (BS 2013) Hadya Elshakh (BS 2013) Mohammad Younes (BS 2013)

Faculty Advisors Annette Vincent, Ph.D.

Category Biological Sciences

Abstract: Genetically modified organisms (GMOs), have their genetic makeup modified to confer a specific trait that allows for an advantage in growth. To date, genetically modified corn, soy and canola are widely available in the market. Although GMOs are not grown in Qatar, it is unknown whether or not imported ingredients used in the manufacture of various food products contain GMOs. The purpose of this project was to determine the presence of corn and soybean in various food products. This was done using both DNA and protein based analysis. Single and multiplex PCR were conducted for DNA analysis, to detect the presence of various transgenic modifications that are common in corn and soybeans. The most common promoter and terminator, CamV35S and NOS respectively were targeted for the soybean. Furthermore, the most commonly inserted genes, namely pat and CP4 EPSPS, were also targeted for the soy. For the corn the invertase, pat, bar and cry9B genes were targeted during PCR, as well as the CamV35S promoter. However, the PCR showed no apparent indication of GMOs in any of the food products. For the protein-based analysis, an ELISA was used to detect the PAT and CP4 EPSPS enzyme products of the pat and CP4 EPSPS transgenes in GMOs. As an ELISA is a highly sensitive detection method, the CP4 EPSPS was detected in all food products indicating the presence of GM corn and soybean in terms of the CP4 EPSPS transgene. However, the PAT enzyme was not detected suggesting that the pat gene was not used in constructing the GM corn and soy in the food products. The results of the experiment suggest that the food products contain GMOs, specifically the CP4 EPSPS transgene was used in producing GM corn and soybean. Further analysis needs to be done to confirm the results.

7


Detection of Genetically Modified Organisms in Food Products Hadya Elshakh

helshakh@qatar.cmu.edu

Hiba Al-Ashtal

hashtal@qatar.cmu.edu

Introduction

http://www.hiddensoy.com/soy-health-risks.php

Mohammad Younes myounes@qatar.cmu.edu

Annette Vincent, Ph.D annettev@qatar.cmu.edu

Purpose

http://www.agmrc.org/commodities__products/vegetables/sweet-corn/

• GMOs are organism’s whose genetic makeup is altered to confer a desirable trait. Currently, most GMOs are crops that contain growth advantages such as herbicide resistance. • A cassette with a promoter, transgene, and terminator is inserted into the host genome’s DNA by various methods. • Soy and corn are the main crops that are genetically modified. Although Qatar does not grow GM crops, it is unknown whether imported ingredients contain GMOs

• Producing GMOs has been an active area of research since the 1990s. • Researches have been searching for new methods and advantageous transgenes to make GMOs and enhance crop yield. • The downfall of GMOs is that a certain modification or spontaneous mutation may produce allergens or other unknown consequences. • Hence, consumer food products containing GMOs must be labeled. Labeling requires the ability to detect genetic modifications. • In this study, GM soy and corn were attempted to be detected in food products

[1] James, D., Schmidt, A., Wall, E., Green, M., Masri, S. 2003. “Reliable Detection and Identification of Genetically Modified Maize, Soybean, and Canola by Multiplex PCR Analysis”. Journal of Agricultural and Food chemistry. 51: 5829-5838. http://www.21food.com/products/soya-beans-and-soya-bean-oil-419294.html

Methods

• There are two approaches to detect the presence of GMOs. First, DNA analysis can be conducted to detect the inserted transgene. Second, protein based methods can be used since usually transgenes will encode for a protein • The Polymerase chain reaction (PCR) is used to amplify a specific DNA sequence such that it can be detected. Multiplex PCR was used as it targets multiple genes for screening within a single reaction. 1. Soy PCR targets: CamV35S promoter, Nos terminator, Lec1, pat, cp4 EPSPS 2. Corn PCR targets: CamV35S promoter, Nos terminator, invertase, bar, cry9b, pat, cp4 EPSPS • An Enzyme-Linked Immunosorbent Assay (ELISA) was used to detect proteins of interest. PAT and CP4 EPSPS are commonly grown soy crops as they confer resistance to LibertyLink® and Roundup Ready® herbicides respectively.The PAT and CP4 EPSPS enzymes were targeted via a sandwich ELISA.

http://en.wikipedia.org/wiki/Kit_Kat

http://www.americansweets.co.uk/quaker-yellow-corn-meal-mix-canbe-used-for-nigellas-spoonbread-680g-625-p.asp

http://www.21food.com/products/corn-curls-processing-line-84301.html

Results Figure 1: Multiplex PCR Gel Electrophoresis Results for Soy and Corn Food Products.

Table 1: ELISA Results for PAT and CP4 EPSPS in Soy and Corn Food Products.

http://xnet.rrc.mb.ca/davidb/other_nucleic_roles.htm

Conclusion

http://www.leinco.com/sandwich_elisa

• Multiplex PCR was inconclusive, likely due to non optimal conditions such as low DNA concentrations • Roundup Ready soy grown locally but notlabeled • Could’ve been PAT in KitKat, cornmeal, and crispy curls chips but it was too processed • Imported ingredients may not be tested for GMOs


Investigation of Legal and Regulatory Obstacles to Becoming an Entrepreneur in Qatar Authors Chaudry, Anas (BA 2013) Shahid, Raheem (BA 2014)

Advisor George White, Ph.D.

Category Business Administration

Abstract The poster will demonstrate what we have learnt through our research on legal and regulatory obstacles to becoming an entrepreneur in Qatar. It contains information related to what we gathered from our surveys, interview and literature research. It is a summary of what we have concluded from the research and some background on the topic. Qatar is a prosperous and growing nation; and has recently become the richest economy in the world in terms of GDP per capita. It has opportunities for investors and entrepreneurs. However, there is a constraint on non-Qatari entrepreneurs who want to open a new business presented by the Qatari ownership law. This could restrict their aspirations and many ideas could be left there in the void. This research examined how the ownership law may discourage non-Qatari ownership and could hinder economic growth. Primary and secondary research will be used to answer this question. Primary research included field surveys and secondary research included an evaluation of the literature. Analyzing the graduating class, their aspirations to become entrepreneurs, to become leaders and the challenges they faced and other stories related to startups in Qatar, we came up with this topic. Furthermore, Qatar vision 2030 tries to empower its citizens, so we wanted to see what laws exist to increase entrepreneurship in Qatar. There are misconceptions related to entrepreneurship laws operating in Qatar, which we wanted to change amongst the society. Moreover, this step was necessary because although the information is there, no compact source containing such knowledge is present, which could be useful for aspiring entrepreneurs.

9


Investigation of Legal and Regulatory Obstacles to Becoming an Entrepreneur in Qatar

How does the ownership law discourages ownership and hinder economic growth in Qatar? Anas Chaudry

Raheem Shahid

Introduction:

• Qatar is an oil & natural gas based economy •Second highest GDP per capita •Qatar 2030 vision: Business oriented & knowledge based economy •People prefer jobs over entrepreneurship

Population Breakdown Qataris 16%

Arabs 13%

Expats 71%

Advisor: George White

Problems:

•Under ownership Law 13 of 2000, an expat can only own 49% of the business •Amendments to the Law in 2010 allowed 100% ownership in all sectors except banking, insurance and real estate •There is an apparent miscommunication of this law •There are several that an expat is required to go through before completing the process •Therefore, fully owned business becomes difficult for an expatriate Questionnaires

20 Surveys Entrepreneurs

One-on-one Interviews

12 Interviews of Qatari & Non Qatari Entrepreneurs

Primary Research

Methodology Internet Articles Literature Review Government Data

Research Results:

•Incentives are misaligned with entrepreneurship •Mutual agreement of profit sharing •Maturity of the business ecosystem is not present •Helps to retain knowledge in the country •Encourages the expatriates to follow Qatar 2030 vision

Conclusion:

Age Group

Gender

• 18 – 40 years

• Male • Female

Findings:

Nationalities • Qataris • Non Qataris (Including other Arabs)

•25% of the expats did not pursue the business because of the law •70% Qataris did not like the idea of the ownership law •90% expats resented the law

•Qatar has taken initiative to become a knowledge based economy with the establishment of Qatar 2030 vision •People face regulatory problems pertaining to entrepreneurship •It will take time for the incentives and the environment to align for the benefit for the entrepreneurs.

9 8 7 6 5 4 3 2 1 0

Would it be good for the Qatari economy if the government eliminated ownership law?

Yes

No

Don't care


Home-Based Business: A Growing Phenomenon Authors Abdulrahman Al-Muftah (BA 2013) Maryam Al-Subaie (BA 2013) Maryam Al-Thani (BA 2013)

Faculty Advisors George White, Ph.D.

Category Business Administration

Abstract During the last five years there has been an upsurge in the number of home-based businesses in Qatar. This research investigates the reasons behind the sudden surge in home businesses, the support they receive from organizations such as Bedaya, Social Development Center and Roda Center, as well as the future of home-based businesses in Qatar. Many young entrepreneurs in the country have noticed an absence in the market for businesses that sell desserts (cupcakes more specifically), fashion accessories and abayas. Creating a home-based business was one way of seizing this opportunity and meeting the market demand, at minimal start-up costs and reduced risk. Interviews were conducted with three centers that provide career and entrepreneurship guidance as well as seven home-based businesses that have been operating for at least a year. The research showed that Bedaya, Social Development and Roda Centers played predominant roles in the rise of popularity in this type of business. The advancements in social media have also made it very easy for aspiring entrepreneurs to reach out to customers at relatively low costs. The growth of home-based businesses in Qatar is an undocumented phenomenon. The study increases the awareness of home-based business in the country and is a research platform for further investigation in the field.

11


Home-Based Business: A Growing Phenomenon Faculty Advisor: George White, Ph.D.

Maryam Al-Subaie | Maryam Al-Thani | Abdulrahman Al-Muftah

Introduction

The aim of this study is to investigate the success and growth of home based businesses in Qatar. The phenomenon of home businesses in Qatar is a result of a lack of desert shops and abaya shops in the region. At the time when these home based business owners started creating cupcakes there were very few or even no cupcake stores in Qatar whereas cupcakes are very popular in the UK and the USA. Many residents in Qatar were fascinated by the cupcakes sold abroad and these cupcake owners realized the incredibly high potential for cupcake stores in Qatar The cupcake and desert home businesses in Qatar are very successfu The same methodology applies to abaya designers, there are two to three main abaya stores in Qatar that are sold in the malls in Qatar and they are Al-Motahajiba, Hanayeen and My Fair Lady. However, these styles in the eyes of many consumers lack creativity and originality and are pro-

Research

1. Entrepreneurship & Career Guidance Centres:

Roudha Almost %60 of home businesses in Qatar, are started up by women; a statistic Roudha and Carnegie Mellon could not ignore. This center would become the first of its kind to not only help out struggling home businesses, but also most importantly legalize them. 36%

40%

42%

28%

39%

38%

38%

26%

1972

1977

1982

1987

1992

1997

46%

Terzi Designs: abayas

Other 20% Government 17%

step 1 Innovate

50%

Roudha’s main focus, are home businesses in Qatar. Having said this, the rise in female students in Qatar outnumbers the number of male students. This, has resulted in more businesses opened up by women, as this clever illustration depicts below: Barriers/ Challenges

Bedaya When approached by entrepreneurs and homebusinesses, Bedaya usually follows a 5 steps process.

step 5 Launch

46%

Kharaweesh: handmade accessories

Family 2%

Budget 41%

Aishaz Collection: abayas

Business Knoweledge 20%

2. Home-Based Businesses: Case Studies from Doha step 2 Design

step 4 Fund

Remeya: photography studio

Results & Conclusions

step 3 Promote

Social Development Center The Social Development Center launched a program,“Badr wa int Gadr” which was established in 2000 in order to help and provide full entrepreneurial support to entrepreneurs in Qatar. The primary objectives of the “Badr wa int Gadr” program are to increase the success rate of entrepreneurial projects and to develop the entrepreneurship environment in Qatar. of their family in terms of business experience and knowledge.

Frosting Qatar: bakery

In conclusion, there are two problems that home business owners and potential home business owners face. The first problem involves the reluctance of some Qatari women to present in front of, or, interact with men. The second problem that these home business owners face is that they are not aware of the importance of legalizing their business.


An Authorization Model for the Web Programming Language Qwel Author Lulwa Ahmed El-Matbouly (CS 2013)

Faculty Advisor Thierry Sans, Ph.D.

Category Computer Science

Abstract With the fast growth of web technology, it is becoming easier for developers to design and deploy complex web applications. However, securing such web applications is becoming an increasingly complex task as this technology provides limited support. One of the main challenges is to be able to restrict access to specific users with the right privileges. For example, a university might want to restrict access to internal web services to students currently enrolled in its program. Writing such a secure web application is complex, because developers are required to reason about distributed computation and to write code using heterogeneous languages, often not originally designed with distributed computing in mind, nor built-in security features. Qwel is an experimental programing language for the web that aims at providing developers with an expressive language that is simple and easy to use for the purpose of building web services with embedded security constraints. The goal of this research is to extend Qwel with built-in security features to allow developers to define access control policies and issue credentials to users. At first, we designed an access control logic to reason about distributed access control. Later on, based on this logic, we were able to extend the syntax and semantics of Qwel with new security features. Using this extension, a programmer is able to grant credentials to users and express security policies. These policies are checked dynamically and, only when they are satisfied, access to the requested service is granted. As a result, we showed how this extension can be used to express common scenarios as well as more sophisticated ones that are beyond current technology.

13


An Authorization Model For The Web Programming Language Qwel

Problem

Lulwa Ahmed El-Matbouly

lulwa@cmu.edu

Thierry Sans (Advisor)

tsans@qatar.cmu.edu

Securing web applications is a complex task

With the fast growth of web technology, it is becoming easier for developers to design and deploy complex web applications. However, securing such web applications is becoming an increasing complex task as this technology provides limited support. One of the main challenge is to be able to restrict access to specific users with the right privileges.

Solution

Designing a better programming language for building secure web application

Qwel is an experimental programing language for the web that aims at providing developers with an expressive language, yet simple and easy to use for the purpose of building web services with embedded security constraints. The goal of this research is to extend Qwel with built-in security features to allow to developers to define access control policies and issue credentials to users. Calculates the similarity report and returns it with the original document Sends the document to noplagiarism.com and obtains the similarity report in return similarityReport@noplagiarism.com

Plagiarism

let simReport = call similarityReport@noplagiarism.com with “Once upon a time ...” in call submit@submission.org with simReport end

1

publish doc:string = let report = <doc, calculateSimilarity(doc)> in report end

2

submit@submission.org

publish s as <doc,report> = store(s)

Forwards the document and its similarity report to the submission.org

Methodology

Stores the original document and its similarity report

Extending Qwel with new features to express credentials and access control policies

First, we designed an access control logic to reason about distributed access control to be able to

·express credentials as properties issued by trusted authorities e.g. the similarity report comes from noplagiarism.com ·express security policies as sets of conditions that must be true based on issued credentials e.g. only a student from univ.edu can obtain a similarity report

Second, based on this logic, we were ale to extended the syntax and semantics of Qwel with new security constructs. Creates a credential vouching that plagiarism.org is the issuer of the report

Creates a credential vouching that the caller (Alice here) is a student at univ.edu

similarityReport@noplagiarism.com

publish caller doc:string = let report = <doc, calculateSimilarity(doc)> plagCred = cred(issuer(report)) in <report,plagCred> end protect univ.edu says student(caller)

login@univ.edu

publish caller p as <login,pwd> = if check(login,pwd) then cred(student(caller)) else raise AccessDeniedException

0 Sends the login and password to univ.edu and obtains a student credential in return

Plagiarism 1 2

let univCred =

call login@univ.edu with <“Alice”,”pass4alice”> <simReport,plagCred> = call similarityReport@noplagiarism.com with <“Once upon a time ...”,univCred> univPlagCred = univCred + plagCred

in

call submit@submission.org with <simReport,univPlagCred> end Concatenates the univ.edu credential with the noplagiarism.com credential

Conclusion

The policy ensures that univ.edu issued the credential saying that the caller (Alice here) is a student

submit@submission.org

publish caller s as <report,cred> = store(s) protect univ.edu says student(caller) and noplagiarism.com says issuer(report)

The policy ensures that: 1) univ.edu issued the credential saying that the caller (Alice here) is a student and 2) noplagiarism.com issued the credential saying ensuring its authenticity

A web language with primitives for access control

Using this Qwel extension, a programmer is able to grant credentials to users and express security policies. These policies are checked dynamically and only when they are satisfied, access to the requested service is granted. As a result, we showed how this extension can be used to express common scenarios as well as more sophisticated ones that are beyond current technology.


Arabic Accented Facial Expressions for a 3D Agent Author Amna AlZeyara (CS 2014)

Faculty Advisors Majd F. Sakr, Ph.D. Micheline Ziadeh

Category Computer Science

Abstract In this study we attempt to identify and develop a suitable set of facial expressions for a 3D agent that has Arabic facial features. We also evaluate how Arabs and non-Arabs recognize Arabic facial expressions implemented on a 3D agent. As shown in previous research, a 3D agent’s emotions enhance the interaction between the agent and humans. The expressions displayed along with the emotion should match human behavior, as people apply their social rules on the 3D agents. Expressions, as in language, exhibit culture-specific accents when expressing certain emotions. Hence, expressions should relate to a specific culture, otherwise, misunderstandings might occur when interpreting facial expressions. Furthermore, people expect a 3D agent’s behavior to align with its features. Therefore, a 3D agent with Arabic features must display Arabic accented facial expressions. In this work, we implement six Arabic facial expressions on the 3D agent based on Arabic human facial expression. A database of Arabic accented facial expressions does not exist. To overcome this, we acquire expressions by taking videos of Arab females narrating stories that invoke different emotions. We extract the expressions from the videos and implement them on the 3D agent using the Facial Action Coding System. We then run a survey to evaluate Arabs’ and non-Arabs’ recognition of the facial expressions. Our results show that the recognition of positive expressions (happiness and surprise) is high and differ slightly between Arabs and non-Arabs. As for negative expressions (anger, disgust, and sadness) the recognition is low and also differs slightly between Arabs and non-Arabs, except for disappointment, which is better recognized by Arabs. These results reveal that positive expressions, when implemented on a 3D agent, are easily recognized. Further, we find that disappointment is a culturally dependent expression for Arabs.

15


Amna  AlZeyara   amna@cmu.edu  

Advisors:  Dr.  Majd  Sakr   msakr@qatar.cmu.edu  

Problem  statement  

Mecheline  Ziadee   mziade@qatar.cmu.edu  

Overall  Goal  

•  Facilitate  the  interac,on  between  humans  and   a  3D  agent  by  focusing  on  the  development  of   the  3D  agent’s  facial  expressions   •  Evaluate   Arabs   and   non-­‐Arabs   recogni,on   of   Arabic  expressions  implemented  on  a  3D  agent  

Improve  the  interac,on   between  humans  and  a   cross-­‐cultural  3D  agent  

Mo&va&on   Humans   expect  agents   to  show   human-­‐like   behavior  

A  3D  agent   must  display   emo,ons  to   be  believable    

The  3D  agent   features   induces   opinions   about  it  

Approach  

A  3D  agent   with  Arabic   facial  feature   must  display   Arabic  facial   expressions  

The  experiment  

Collec&on  

1)  Select  6  emo,ons:  anger,  disgust,  surprise,   happiness,  sadness  and  disappointment.     2)  Iden,fy  the  Arabic  accent  in  each  expression  that   displays  the  emo,on  by  recording  the  expressions   of  Arab  female  students  

We  asked  par,cipants  about:  age,  gender,  origin   and  countries  they  lived  at.  Then  we  asked  them  to   choose  the  most  appropriate  expression  for  each   emo,on  by  using  the  following  matrix:    

Implementa&on  and  Verifica&on  

1)  Implement  the  expressions  using  the   Facial  Ac,on  Coding  System   2)Verify  the  implementa,on  of  the  expre-­‐ ssions  using  a  small  scale  experiment  

Evalua&on  of  recogni&on  

1)  Evaluate  the  recogni,on  of  the  expressions  by   Arabs  and  non-­‐Arabs  using  a  large  scale  experiment  

Conclusion  

Results   Arabs  and  non-­‐Arabs  recogni,on  of  an  emo,on  is   shown  below.  For  each  pair,  Arab’s  percentage  is   shown  on  the  top.     Recognized as

   

Sadness Sadness

Intended emotion

Disgust Anger Disappointment Surprise Happiness

Disgust

Anger

Disappointment

Surprise

Happiness

9%

9%

49%

33%

0%

0%

8%

10%

42%

39%

0%

0%

51%

35%

7%

7%

0%

0%

31%

31%

15%

23%

0%

0%

2%

42%

35%

2%

19%

0%

1%

51%

35%

1%

11%

0%

33%

7%

5%

56%

0%

0%

59%

4%

0%

35%

1%

0%

2%

5%

2%

0%

74%

16%

0%

3%

3%

0%

79%

15%

2%

2%

2%

2%

7%

84%

0%

1%

4%

1%

8%

85%

•  • 

• 

Arabs  and  non-­‐Arabs  beRer  recognize  Arabic   posi,ve  facial  expressions  that  are   implemented  on  a  3D  agent.   Although  sadness  and  disappointment  are   claimed  to  send  the  same  signal,  Arabs  were   beRer  able  to  dis,nguish  between  the  two   expressions.       Disappointment,  when  implemented  on  3D   agent,  is  more  culturally  dependent  than   other  expressions.  


CMUQ Official Android App Authors Ali Elgazar (CS 2015) Danah Abdullah (CS 2015)

Faculty Advisors Mark Stehlik

Category Computer Science

Abstract The poster serves to promote the Carnegie Mellon University in Qatar (CMUQ) Android Application, which is currently under production and has an existing functional demo. Though the poster does not detail all the functions of the application, an image of the phone with the application running will illustrate a general concept of its capabilities. The research topic involves the development of the first ever CMUQ official android application, which serves as a database of information regarding current news and events taking place at CMUQ. The application will be made available on the market for students. The inspiration for the app came from a personal experience. One of the authors was inspired to develop the application during winter break, when he missed an event that he had learned about too late. At CMU-Q, students are often bombarded with so many events that it becomes tedious to keep track of all of them. The app is designed to remedy the situation. The application needed to be quickly and easily accessible, so designing it for PC was ruled out in order to avoid the hassle of logging in whenever a user needed to check for upcoming events. The best solution would be to have the application with the user at all times. For that reason, we decided to develop the application for android, using eclipse and a Samsung S3 phone during the research and testing process. This research project is of great importance for the local community.

17


Interaction Analysis of a Multi-Lingual Robot Receptionist Authors Naassih Gopee (CS 2016) Lamana Mulaffer (CS 2015) John Naguib (CS 2016) Micheline Ziadee

Faculty Advisor Majd F. Sakr, Ph.D.

Category Computer Science

Abstract: In this work we analyze human interactions with Hala, a bi-lingual robot receptionist, in order to identify variations across Arabic and English interactions. The larger scope of this work is to study the influence of socio-cultural norms on human-robot interactions within a multicultural, yet primarily ethnic Arab, setting. Our analysis will help impact future design decisions to better adapt the robot to interact in those specific languages. In previous work we have identified several variations in user interactions with the robot. Specifically and for English interactions, we have observed that native Arabic speakers converse 25% longer than non-Arabic speakers and they tend to thank the robot less than native English speakers. Hala has been deployed at Carnegie Mellon University in Qatar’s reception since October 2009. Hala utilizes a rule-based matching system in order to respond to user queries. The number and spectrum of rules dictates Hala’s ability to provide relevant answers to a wide variety of queries. The majority of Hala’s rules (96%) were in English until December 2012. After that point, we developed symmetric content between English and Arabic rules. All interactions with Hala, whether in English or Arabic, are logged. In order to identify if variations exist we analyze the logs. A script is developed to write the logs into a MySQL database and the log data is thereafter cleansed to remove spurious interactions. Statistical analysis is performed on the data in order to identify variations of interactions across different languages. According to our analysis, 85% of the interactions are in English, 3% in Arabic and 12% are mixed (Arabic and English) interactions. The average duration of an interaction across languages is under 2 minutes and includes an average of 4.73 queries. However, English interactions contain twice as many queries as Arabic ones and mixed interactions contain more than twice as many queries as English interactions. After equalizing Hala’s English and Arabic knowledge-base, we observe an increase in mixed interactions by 6.5%. As for the Arabic interactions, the average duration is 8% longer and the average number of queries is up by 25%. These results indicate significant variations in human-robot interactions across different languages. Increasing the Arabic content has impacted the Arabic interactions. In our future work, we attempt to further study these cross language variations in multi-lingual human-robot interaction.

19


I Want my Mommy Authors Fahim Dalvi (CS 2014) Syed Moosavi (CS 2014)

Faculty Advisor Saquib Razak, Ph.D.

Category Computer Science

Abstract “I want my Mommy” is a research project that looks at using wireless technologies such as Bluetooth to solve the problem of crowd control by limiting the number of lost people in crowds. The concept of the solution is based on the fact that in crowded areas, people want to stay with their groups. In order to make sure that people stay with their groups or have information about the location of their group members, we develop the algorithm that allows a device to automatically detect one’s group members and then provide the location of these group members in case that person is lost. In this research project, we would like to find out how effective our algorithm is in finding groups within a large crowd and how helpful the device is in helping lost people. For this research project, we would like to simulate the situation of a crowded area and give devices to a number of students that would want to participate in the research study. After providing the students with the devices, we will check after a few days if the devices have selected the correct group members. We would then see if a device would be able to provide the location of the group members when the device is not in proximity with other devices.

21


I want my Mommy Using Wireless Research to limit the number of lost people

Abstract

Fahim Dalvi

Syed Hashim Moosavi

Saquib Razak

fid@qatar.cmu.edu

syedhashim@cmu.edu

srazak@qatar.cmu.edu

“I want my Mommy ” is a research project that aims to use wireless technologies such as Bluetooth and Wi-Fi to quickly locate people in a large crowd, subsequently reducing the number of lost people. In several crowded areas such as Makkah and Disneyland, people getting separated(specially children and elderly) from their families is a huge problem. This is currently handled manually by making announcements or giving people tags with information written on them. Unfortunately, these solutions do not work in highly crowded areas, both because of the number of people entering the location, and because of the size of these places. We plan to devise an algorithm using commonly existing wireless technologies to reduce the number of lost people by categorizing the crowd into groups without any barrier-to-entry.

Problem

• Crowded places generally have a lot of people entering the premises • Identifying groups at this point is time consuming, and is a barrier to entry

• People spend most of their time with their respective groups • The time spent with the group initially is not used for the classification of an individual’s “neighbors”

Solution

• A lost person is unable to locate his/her group in the large crowd due to lack of infrastructure • Manual announcements/ non-electronic tagging is used that is not very effective

Experiment

• Identify an individual’s “neighbors” based on the proximity in the crowded area and the time of contact by using Bluetooth • Use the Wi-Fi of the individual’s wireless device (Android phone in our experiment) to get the latest location of the individual and periodically send it to the server • Find the best way to categorize people into groups by using the information from the wireless device and provide them with the way of reaching out to them when lost

• Android phones are being used by 20 students of CMUQ for Wireless research and their wireless data is continuously being collected for analysis • Both the location (with respect to the access points) and the proximity of an individual to his/her neighbors helps in the classification of groups • The experiment will be performed with a larger group of students after preliminary results to simulate a crowded area


Semiotic Circles: An Inclusive Methodology for Human Computer Interaction Design Author Kenrick Fernandes (CS 2014) Faculty Advisor Andreas Karatsolis, Ph.D

Category Computer Science

Abstract With the proliferation of increasingly powerful and portable computing devices throughout our lives, the quality of our experience using these devices becomes ever more important. Usersâ&#x20AC;&#x2122; interactions must be supported, not controlled, by these devices. In the field of human-computer interaction (HCI), the use of human-centered research methods is essential for a more inclusive user experience. Our projectâ&#x20AC;&#x2122;s objective was to design a methodology which would help us obtain participantsâ&#x20AC;&#x2122; perspectives regarding an educational symposium. Helping these ideas surface enabled us to work towards the creation of a web presence to support the symposium. The hybrid methodology we designed is based on existing paper prototyping and think-aloud protocols with the addition of an offline tagging system. This combination of tools, along with the tagging system, supported our creation of composite visualizations to represent a synthesis of viewpoints. We intend to use the knowledge gained from the process to design a web presence that will facilitate and support interactions between people. In this poster session, we present the HCI methodology we created and the protocols we used to move from abstract concepts to concrete tangibles. We also invite you to tag the final composite model and experience the outcomes of this project.

23


Semiotic circleS

An incluSive methodology For humAn computer interAction deSign

student

Kenrick Fernandes CS 2014 kenrick@cmu.edu

faculty advisoR

Andreas Karatsolis English karatsolis@cmu.edu

our ChallEngE

STEP 1 - ThE DESign

Advances in technology have penetrated all aspects of our lives. Consequently, our interactions with devices and other users become more significant and demand meaningful design approaches. In the last few years, several methodologies for hci design have been proposed[1], which attempt to capture information from multi-layered user hierarchies. We have designed an inclusive methodology which acknowledges all participant perspectives and facilitates interaction between stakeholders.

We designed an instrument on a large A0 sized sheet containing concentric circles representing three of the Generalist Questions: Why, How and What ?

our inSPiraTion

STEP 2 - ThE SESSion

Using well-established methodologies such as the Think-Aloud protocol, Paper Prototyping and Object Labeling, we developed a hybrid methodology called Semiotic Circles. This methodology allows for capturing of a participantâ&#x20AC;&#x2122;s thoughts, which we combine with a visualization exercise and tagging system to promote freedom of expression. This way we are enacting a genuinely inclusive model, as all the concepts and relationships between them are represented as participants have identified them.

Participants are briefed about the Think-Aloud protocol and the purpose of the sheet and the tokens. They are then free to place tokens and create connections. The sequence of actions is drawn on the left

Six printed and cut out tokens can be pasted on the visualization sheet. They are used to tag topics and relationships already put down on paper. They represent other Generalist Questions as well as specifying tags for the the web presence and a speech bubble.

STEP 3 - ThE SynThESiS one such synthesized model is below play with it and tag your topics

RefeRences 1. Carroll, J. M. (2003). HCI models, theories, and frameworks toward a multidisciplinary science. San Francisco, Calif.: Morgan Kaufmann. 2. Simon Sinek: How great leaders inspire action | Video on TED.com. (n.d.). TED: Ideas worth spreading. Retrieved April 14, 2013, from http:// www.ted.com/talks/simon_sinek_how_great_leaders_inspire_action. html

Multiple sheets are then synthesized with the tagging tokens as focal points, to generate a 3D model of the aggregate opinions of the group. Tokens are essentially stacked in towers to add an extra dimension.


Multi-Robot Coordination Author Sidra Alam (CS 2013)

Faculty Advosor Saquib Razak, Ph.D.

Category Computer Science

Abstract In the context of complex multi-robot systems, consensus-based task allocation algorithms are being researched extensively because of their robustness in handling the growing number of robots in a team [1]. Multi-robot systems are being studied and developed in the context of disaster stricken areas. In such critical scenarios, coordination and communication during task allocation become vital factors to optimally use all the available resources. In this report, we have studied and implemented the consensus-based auction algorithm (CBAA) [2], in parallel with task execution, as proposed by Das et al. in [1]. Furthermore, we aim to propose an algorithm to maintain connectivity in this consensus-based algorithm.

25


Abstract In the context of complex multi-robot systems, consensus based task allocation algorithms are being researched on extensively because of their robustness in handling the growing number of robots in a team [1]. Multi-robot systems are being studied and developed in the context of disaster stricken areas. In such critical scenarios, coordination and communication during task allocation become vital factors to optimally use all the available resources. In this report, we have studied and implemented the consensus-based auction algorithm (CBAA) [2], in parallel with task execution, as proposed by Das et al. in [1]. Furthermore, we aim to propose an algorithm to maintain connectivity in this consensus-based algorithm.

Current Algorithm, Problem & Our Strategy

Experiments We have simulated CBAA in java and have tested it on scenarios with and without range of network communication. We have compared the results to measure performance and to evaluate scenarios where the algorithms produce lower performance.

Results

We are working on the Consensus Based Allocation Algorithm (CBAA) which uses a decentralized approach to delegate the tasks amongst the robots. Agents place bids on tasks with values based on their own Situational Awareness. The highest bid wins the assignment. The testing for this algorithm assumes connectivity at all times which could result in redundant task allocations if implemented on a real system which has a range of communication and hence is prone to network dis-connectivity. In our approach we take in to consideration, network dis-connectivity during task allocation and assign a very high cost to bids which result in dis-connectivity.

Why Do We Care? Multi-Robot Coordination algorithms can be emulated on real mobile systems to deliver the coordination functionality in the event of task delegation. Consider the rescue scenario, where doctors have to attend to patients after a disaster. The doctors play the role of robots in such a scenario and use a mobile device in order to communicate and coordinate task delegation. Time is of the essence and any redundant task allocation could cost a personâ&#x20AC;&#x2122;s life

Future Work

References [1] G. P. Das, Member, IEEE, T. M. McGinnity, Senior Member, IEEE, S. A. Coleman, and L. Behera, Senior Member, IEEE (2011) A Fast Distributed Auction and Consensus Process Using Parallel Task Allocation and Execution [2] H.-L. Choi, L. Brunet, and J. P. How. Consensus-based decentralized auctions for robust task allocation. IEEE Trans. on Robotics, 25 (4):912 â&#x20AC;&#x201C; 926, 2009.

Currently, we are implementing and testing CBAA in the context of network connectivity, in simulation. Next, we would be implementing our proposed protocol to compare it against the existing one. Real-systems: The main aim of this research is to emulate this modified CBAA on a real mobile system and evaluate it.


Towards Computational Offloading in Mobile Clouds Author Afnan Fahim (CS 2013)

Faculty Advisor Khaled Harras, Ph.D.

Category Computer Science

Abstract It is common practice for mobile devices to offload computationally heavy tasks to a cloud, which has greater computational resources. However, this type of offloading has been expensive due to high energy costs and high latency, which exists between the cloud and the offloading mobile device. As an answer to this problem, â&#x20AC;&#x153;Cloudletsâ&#x20AC;? were proposed: smaller clouds placed closer to users that would make mobile task offloading less expensive in terms of energy waste and time consumption. The idea of reducing communication costs by executing closer to the offloader device was then extended to introduce mobile cloud computing - where the idea is to offload tasks to nearby devices - be they mobile or stationary - so as to reduce communication costs and latency. In this work, our first contribution is to motivate the gain in computation time and energy consumed that can be made by offloading to nearby devices. We do this by emulating network conditions that exist for different communication technologies provided by modern mobile devices. As opposed to emulated experiments, currently it is a challenge to carry out real experiments on mobile devices since no applications exist that can distribute different amounts of data and computation to nearby devices using different communication technologies. Thus, as a second contribution, we also present an API that allows creation and offloading of tasks by a mobile device to a network of nearby devices. Another challenge in mobile cloud computing is to decide which device to offload a task to. As a third contribution, our work also presents algorithms to leverage social context when making offloading decisions to nearby devices.

27


Towards Computational Offloading in Mobile Device Clouds Afnan Fahim, Abderrahmen Mtibaa and Khaled A. Harras Carnegie Mellon University Qatar

I. MOTIVATION

II. OFFLOADING OPTIONS

• Mobile usage is overtaking laptop/desktop usage [6] • Average Household in Qatar owns 3 mobile phones and a tablet [5] • New applications require a lot of computation and data processing: • Face Recognition • Video Games • Offloading mobile computation and data is a trend that has just begun • This offloading impacts application latency and consumes energy

III. MAKING THE CASE FOR MDC OFFLOADING Task: Combination of Data (MB) and Computation (MFLOP) Computation: Carried out as floating point matrix operations. Data: Text of known size in MB. Metrics: Offloader Energy & Total Time to Complete Task

• • • •

Client

Server

Java App

Data

Java App

Traffic Shaper

B/W

RTT

Computation

PLR

Java Sockets

IP Firewalls

Computation

Ubuntu

FreeBSD

FreeBSD •

Emulation setup for testing energy and time tradeoffs when offloading

• •

MDC can save both energy and time depending on task complexity and data WiFi based offloading - 2x and 3x faster than Bluetooth and 3G Bluetooth 4 - 80% less energy compared to WiFi

IV. THE MDC EXPERIMENTAL PLATFORM Apps that offload custom tasks do not exist. We thus created an API with these features:

Energy Circuitry

• PreOffloader: Define Tasks as [Data, Computation] • LocalExecutor: Execute tasks on the local device • BluetoothOffloader: Offload tasks in parallel using Bluetooth • WiFiDirectOffloader: Similar to Bluetooth offloader • RemoteOffloader: Offload tasks to remote devices using WiFi

Data vs Energy - 30 MFLOP

Energy (J)

70

Total Energy

60

Computation Energy

50

Local Computation

30 20 10 0

Total Time

120

Computation Time

100

60 40

0

Computation Time

100

Local Computation

90 80

80

70

60 40

20

20 0 0

5

10 Data (MB)

20

5

10

60

Total Energy Computation Energy Local Computation

50 40 30

0

5

10 Data (MB)

20

20 10 0

0

5

10 Data (MB)

References

1. 2. 3. 4. 5. 6.

20

Data vs Energy - 60 MFLOP

Total Time

Energy (J)

Local Computation

Time (s)

Time (s)

80

0

Data (MB)

140

100

V. CONCLUSIONS & ONGOING WORK

40

Data vs Time - 60 MFLOP

120

MDC App

80

MDC Experimental Results Data vs Time - 30 MFLOP

Energy Results

20

• For compute intensive tasks we have shown energy and time gain using MDC • We can now test different offloading strategies using the MDC API • Identifying when and who to offload computational tasks using social context and contact history • Addressing intermittent connectivity in MDC devices

C. Shi, V. Lakafosis, M. H. Ammar, and E. W. Zegura, “Serendipity:enabling remote computing among intermittently connected mobile devices,” in MobiHoc, 2012, pp. 145–154 B.-G. Chun, S. Ihm, P. Maniatis, M. Naik, and A. Patti, “Clonecloud: elastic execution between mobile device and cloud,” in Proceedings of the conference on Computer systems, ser. EuroSys ’11. New York, NY, : ACM, 2011 E. Cuervo, A. Balasubramanian, D. ki Cho, A. Wolman, S. Saroiu, R. Chandra, and P. Bahl, “Maui: making smartphones last longer with code offload,” in MobiSys’10, 2010 Mahadev Satyanarayanan, Paramvir Bahl, Ramón Caceres, and Nigel Davies. 2009. The Case for VM-Based Cloudlets in Mobile Computing. IEEE Pervasive Computing 8, 4 (October 2009), Qatar’s ICT Landscape 2013: Households and Individuals – ICT Qatar Annual Report - http://www.ictqatar.qa/sites/default/files/documents/Qatar%20ICT%20Landscape_EN.pdf Microsoft Tag Mobile Usage Survey - http://www.digitalbuzzblog.com/2011-mobile-statistics-stats-facts-marketing-infographic/


PyExoplanets: A Computed Application for Detecting Exoplanet Transits on Stars’ Light Curves Authors Noora J. Al-Muftah (CS 2016)

Faculty Advisor Saquib Razak, Ph.D.

Category Computer Science

Abstract

One way of detecting extrasolar system planets (planets that orbit stars outside our solar system) is by the transit method. In this method, astronomers use the light curves (plot of the brightness of a star over a period of time). If a decrease in the brightness of the star is noticed in the light curve, then it can be further deduced if it is caused by an object (planet) that blocks the star’s emitted light by passing in front of it. Projects like NASA’s Kepler mission produces thousands of stars’ light curves searching for exoplanets. One resulting problem is that it is difficult and time consuming to analyze all of these light curves one by one. For my project, I have developed an astronomical data analysis application that can shorten the process of finding transits by applying an image analysis method on the graph of the light curve. In my application, PyExoplanets, I selected light curve data from the Kepler mission FITS files for planetary-candidate stars observed in the Q3 quarter set. My application was developed using Python, the Python Image Library for analyzing the light curve image, PyFITS for reading FITS files, Matplotlib for plotting the star’s data, and Visual Python for producing simple 3D animation of the star and the exoplanet. My project is an initial step in creating a complete computer application that can potentially assist astronomers in the process of finding exoplanets.

29


PyExoplanets  

 Computed  Application  for  Detecting  Exoplanets’  Transits  on  Stars’  Light  Curves    Noora  J.  Al-­‐Mu-ah  

       nmu%ah@qatar.cmu.edu              

Problem  and  Motivation   •   

 Current  projects  such  as  NASA’s  Kepler  mission  [1]    produce  thousands  of  stars’  light  curves  searching    for  extrasolar  planets  

•   

 DetecDng  Exoplanet  transits  on  large  light  curve  data    set  takes  Dme  and  effort    

•   

 An  applicaDon  is  needed  to  assist  in  this  process    by    detecDng  possible  planet  transits  and  hence    illuminaDng  the  number  of  false  planetary  candidates  

 

Advisor:  Saquib  Razak,  Ph.D.   Computer  Science  

Main  interface  of  Application  

 [1]  hJp://kepler.nasa.gov/  

 

Solution   Transit  DetecBon  Method:   •     Created  an  Image  Analysis  method  that  detects    Exoplanet  transits  on  light  curves  and  outlines    the  region  

Animation  interface  

•       Algorithm  analyzes  the  data  trends  and  outlines    any  possible  transits  on  the  graph  of  the  light    curve   Data  source:  Kepler  Q3  quarter  set  

hJp://archive.stsci.edu/pub/kepler/lightcurves/  

 

Future  work   PyExoplanets  Application  features   I.  Analyzes  individual  Kepler  planetary  candidates   II.  Plots  staDsDcs  for  a  set  of  Kepler  planetary   candidates   III.      Animates  a  3D  visualizaDon  of  star  and   Exoplanet      

  •   

 Improve  the  Image  Analysis    method  to    increase          accuracy  and  reduce  false    detecDons  

•       Develop  the  applicaDon  to  be  able  to  handle  all    types  of  light  curves’  plots   •                 Enhance  runDme  for  the  detecDon  process  


Unsupervised Word Segmentation and Statistical Machine Translation Author Hanan Alshikhabobakr (CS 2013)

Faculty Advisor Kemal Oflazer, Ph.D.

Category Computer Science

Abstract Translation is highly dependent on how much we know about the text, so the more accurate analysis the word given to a translator, the better the translation result we get. One way of analyzing a word is to segment the word into grammatically meaningful linguistic segments (Such as “playing” segmented into “play+ing”). This research looks at Arabic word segmentation that is independent of human supervision, so we are measuring the precision of Unsupervised Word Segmenters. We then measure how much the Unsupervised Word Segmenter improves translation. This research utilized the servers allocated for the Natural Language Processing group at CMUQ, where they store the state of the art systems of NLP, which require a relatively huge amount of disk space to run them. This research will contribute to the development of intelligent machine translation, especially to the Arabicto-English translation.

31


What Affects Students’ Acceptance and Use of Technology? A test of UTAUT in the context of a higher-education institution in Qatar

Author Fatema Akbar (IS & BA 2013)

Faculty Advisor Daniel C. Phelps, MLIS, Ph.D., CISA

Category Information Systems

Abstract Technology is increasingly being integrated in classrooms to facilitate and enhance students’ learning. However, the success of new technology introductions cannot be achieved if the students do not accept and use the technology. This research project examined the factors influencing students’ acceptance and use of technology in a higher-education institution in Qatar by testing the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et. al, 2003). UTAUT is one of the most prominent theoretical models of technology adoption. It synthesizes eight previous user acceptance and motivation models and suggests that four core constructs are direct determinants of technology acceptance and use: Performance Expectancy, Effort Expectancy, Social Influence, and Facilitating Conditions. The theory also suggests that the effect of these four constructs is moderated by age, gender, experience and voluntariness of use. Among the studies citing UTAUT, very few implement the full model and examine all of its constructs. This study uses the complete UTAUT as a model to examine the factors that influence students’ acceptance and use of technology in a new organizational and cultural context that UTAUT has not been previously tested in. This study can provide recommendations to help prepare the right environment and training before a new technology is introduced for students. Additionally, this research adds to studies about the applicability of UTAUT in an educational environment by extending the diversity of the studied sample.

33


Fatema Akbar IS & BA 2013

fakbar@qatar.cmu.edu

WHAT AFFECTS STUDENTS’ ACCEPTANCE & USE OF TECHNOLOGY?

A test of UTAUT in the context of a higher-education institution in Qatar

INTRODUCTION

METHOD

WHAT?

WHERE?

WHY?

Unified Theory of Acceptance & Use of Technology (UTAUT)

Higher Education Institution in Qatar

To examine factors influencing students’ tech acceptance

UTAUT was tested by conducting a panel study collecting data from students in classes where new technologies were introduced. The students filled online surveys about their thoughts and feelings towards the technology three times throughout the semester to measure how their perceptions change with increased experience.

O X X Introduction/ Student System Reactions Use Training

UTAT is a prominent theory providing a technology acceptance and use model that synthesizes eight previous technology adoption and motivation models. This study uses the complete UTAUT as a model to examine the factors that influence students’ acceptance and use of technology in a new organizational and cultural context that UTAUT has not been previously tested in.

T1: 1 week

O Student Reactions/ Usage Measurement T2: 4 weeks

X System Use

O Student Reactions/ Usage Measurement T3: 8 weeks

Students’ acceptance and use of technologies introduced in their academic environments is an important factor in determining the success of these technologies.

MODEL HYPOTHESES & RESULTS Performance Expectancy

Supported Partially supported Not supported New findings

Effort Expectancy Behavioral Intention (Acceptance)

Social Influence

Use Behavior (Actual Use)

Facilitating Conditions

Attitude towards using technology

Gender

Age

Experience

Voluntariness

CONCLUSION As shown in the graph above, some hypotheses of the UTAUT model were supported, while others were not. The differences in results between this study and what the UTAUT model suggests can be attributed to the different cultural and organizational contexts this study tests UTAUT in.

Other constructs and moderators had a significant influence on technology acceptance only in one point of the semester. For example, computer self-efficacy was significant only in the beginning of the semester when students had no experience with the technology.

The study found that performance expectancy, facilitating conditions and attitude towards using technology are significant determinants of technology acceptance. Additionally, acceptance of technology and facilitating conditions are significant determinants of the actual use of technology.

Future research should focus on adding education context-specific constructs to the UTAUT model and study further relationships between the constructs and technology acceptance. This will help prepare the right environment and training for students before introducing new technologies to improve technology acceptance.


Software Development Project Electronic Resolution Authors Tasneem Jahan (IS 2013) Reham Al Tamime (IS 2013)

Faculty Advisor Maher Hakim, Ph.D.

Category Information Systems

Abstract The Model United Nations, THIMUN Qatar, organizes conferences for high school students where they discuss and debate specific issues by submitting their proposals in document format. The organization uses a document management tool, which is known as a Resolution Management System to organize and manage the documents and resolutions submitted by the students. The system is a critical factor for the success of the conferences, therefore, we have designed a better system that minimizes the errors that have occurred in the past and thus maximizes the convenience in using such a system. Our research focused on finding problems in the current tool, by conducting interviews and focus groups, as well as by observing users. The results contributed in building a new system (Eresolution) by our team that addresses all the current issues. Our vision was to develop a system that is efficient; fulfills the functions of resolution submission and management; saves conference members time, energy and effort; is rich in functions; is user friendly and aesthetically appealing; and is available to use for a minimum amount of charge. The poster will give a summary of the process starting from research, design, implementation, testing and delivering to THIMUN Qatar.

35


Software Development Project Electronic Resolution

Reham Al Tamime: rtamime@qatar.cmu.edu Tasneem Jahan: tjahan@cmu.edu Faculty Advisor: Maher Hakim | Information Systems

Client:

Project Overview

THIMUN QATAR Model United Nations (MUN)

Resolution Management System (RMS):

A system that is used currently by THIMUN Qatar in order to create conferences, committees, and assign users to these conferences and committees depending on the role that they play in the conference. The system also supports different functionalities such as uploading a resolution document, editing it and determining its result after

Needs & Improvement:

Since the RMS is a critical success factor for the MUN conferences, the client was not satisfied about the overall system. Our team got asked to come up with better solution that maximizes the convenience of using such system and solves the current problems.

conference’s debate.

Problems • • • •

Research & Data Collection: • • • •

Current System Problems:

Very Hard to Use Lack Supportive Functionalities Not Visually Pleasing Confusing and Not User Friendly

Observations Test Aloud Protocols Interviews Focus Groups

Current RMS:

E-Resolution:

Solution System Success Attributes: User Friendly • Clear Fonts & Icons • Indicative Arrows for Steps • Interactive Help Page

System Success Attributes: Extra Features Added • • • •

Enter Votes Admin Dashboard & Guest Page Search Features More Account Settings

System Success Attributes: Availability Security Aesthetic Values Added

Implementation Tools: Backend:

SQLight | PostgreSQL

FrontEnd:

Results:

• 100% of sample agree that system is simple and flexible • 80% confirm that the system is very functional and helpful • 90% recognize the easiness of system use and navigation

Testing

HTML | CSS | JavaScript | Bootstrap

Connecting Front & Backend: Ruby on Rails

Deployment:

Heroku | Amazon S3


Privacy and e-commerce in the Arab Culture Author Marwa Al-Fakhri (IS 2013)

Faculty Advisors Daniel Phelps, Ph.D. John Gasper, Ph.D.

Category Information Systems and Social Sciences

Abstract The trend of online shopping is catching on in the Arab world. This research investigates the effect of information availability regarding privacy on Arab consumersâ&#x20AC;&#x2122; decision making when making purchasing attitudes. We hypothesize that the Arab culture will influence the decision making of Arab consumers; the more culturally sensitive the item purchased is, the more consumers are willing to pay for online privacy when presented with information about how private the shopping websites are. Results support this hypothesis, participants exhibited willingness to pay for privacy more on average when they were purchasing cigarettes than when they were purchasing batteries. The findings of this research will give us a better insight into Arab consumersâ&#x20AC;&#x2122; decision making and will better inform the design of online tools that better serve consumersâ&#x20AC;&#x2122; needs.

37


The Effect of type of Display on Conjoint Studies Authors El Houssain El Marabti (BA 2013) Marwa Al-Fakhri (IS 2013)

Faculty Advisor Peter StĂźttgen, Ph.D.

Category Business Administration

Abstract Conjoint analysis is one of the most widely used tools in marketing research to estimate consumersâ&#x20AC;&#x2122; preferences. This research investigates the affect of two factors (display type and ease of data acquisition) on the estimated preferences in conjoint analysis. We hypothesize that the type of display does not affect the relative estimated preferences, but affects the scale of the estimated parameters, as supported by the literature. Also, we hypothesize that attributes that are more easily acquired will be estimated as being more important. Using a 2x2 design, we manipulate display type by representing the choice sets in list format and in pictorial format and manipulate the ease of data acquisition by the number of alternatives in a choice set (3 vs.15 alternatives). Analysis of the results supports both hypotheses. The findings of this research are important for implementation of conjoint analysis studies in the business world.

39


The effect of type of display on conjoint studies

Al-Fakhri abti & Marwa Stuttgen Advisor: Peter

n El Mar rs: El Houssai

Researche

Research question

Does the presentation style and type of display of the different levels of attributes affect the outcome of conjoint studies and ease of info acquisition? Type of Display

Hypotheses Evaluation

Options

Choices

Hypothesis 1 Type of display will affect Ease of Info Acquisition. It will not affect the relative preferences but will affect the scale. Hypothesis 2

number of options

Ease of info Acquisition

Attributes that are easily acquired become more important.

Data acquired

Literature review

(LOUVIERE et al. 1987): Parks study; text vs images. Results suggest no statistical difference. (JANSEN et al. 2009): Housing study; visual vs verbal. Results show directional difference but it was ruled out due to coincidental details. (ORZECHOWSKI et al. 2005): Housing study; text vs multimedia. Results suggest no difference in relative importance of attributes but a different scale.

Methodology Data Collection - Fall 2011

64 participants were recruited in Education City and were compensated with 50 QR. They were shown pictures of the options and their eye movements were tracked to see how many options they looked at before they make their finale choice.

Data Collection - Summer 2012

115 subjects participated in the experiment. They were randomly assigned to one of the three conditions; 3 options text, 3 options images, and 15 options text. Participants were recruited in Education City and were compensated with 36 QR.

Analysis

Multinomial Logit model estimated in Bayesian fashion using MCMC as implemented in the Bayesian package in R.

Estimated relative importances Estimated relative importances of the three attributes in decision making: List of three choices

List of fifteen choices

List of three pictures

List of fifteen pictures

Price, 15% Brand, 26%

Flavor, 59%

Price, 35%

Brand, 27%

Flavor, 38%

List of fifteen pictures different than the other display methods

The graph above shows the different distributions of all the choices display methods, and shows how different the 15 pictures list is from the others.


Arab Spring Newspaper Coverage: A Comparative Analysis Authors Jiyda Mint Moussa (CS 2014) Sophie Qingjia Jiang (NU-Q 2016)

Faculty Advisor David Emmanuel Gray, Ph.D.

Category General Education

Abstract: The Arab Spring has brought a lot of international attention to democratization in the Arabworld. Considering that mass media often is the primary source of information concerning conflicts and revolutions, it is important to examine what type of media coverage has been afforded to events in Middle Eastern and North African nations. In particular, we propose examining and comparing the coverage by Qatari English newspapers against that of newspapers outside Qatar. Did these different newspapers focus on covering the same countries and events? Do they immediately respond to all the important events as they happen? Was there a difference in how they focused on Qatar’s involvement in providing military and financial assistance to certain groups involved in these uprisings? This study examines the similarity and differences in the news coverage of the Arab Spring by comparing the reporting of the New York Times and Guardian with that of Qatar’s Gulf Times and Peninsula. We explore differences in coverage among these four publications in terms of significant Arab Spring events and Qatar’s involvement. The dataset used in this study consists of articles published between December 1, 2010 and December 25, 2012 and retrieved from online archives of the respective newspapers. The majority of the data was gathered via an automated process, for which we designed a computer program to scroll through each newspaper’s archive and extract the articles. Starting with a manually selected collection of articles from different sources, we then created a second computer program to identify whether any given article is likely to be “Arab Spring” related. Using these two programs, we examined the four newspapers’ coverage of the Arab Spring by comparing the number of articles each newspaper published daily in the monitored time frame. We also extracted the most vital Arab Spring events from its timeline and contrast the coverage from those four newspapers. Finally, we analyzed their extents of coverage on Qatar’s involvement in Syria, especially contrasting the distinction between Qatar’s local newspapers and international ones. A breakdown of the number of articles published daily by each newspaper shows that intensity of Arab Spring coverage varied over four consecutive time periods. It also shows that the focus and coverage varied for each newspaper across these different phases. For instance, in the initial phase, the New York Times took the lead with the most published articles while the Guardian and Peninsula followed. However, the Peninsula took the lead in the remaining phases with the highest average number of articles per day. For the Gulf Times, the number of articles was the least throughout and never attained any peak. More importantly, in the first two phases we observed that the non-Qatari newspapers and the Peninsula peak at relatively the same intensity. However, in the last two phases of the reporting, which mostly relate to the Syrian crisis, the Peninsula breaks away from the other two news sources and the amount of coverage radically increases. This last trend may reflect Qatar’s political interest in this crisis, a topic for further exploration. 41


because it took a notable lead with reporting on the battle of Tripoli in August, 2011.

Finally, Guardian had a special interest in covering Libya, not only because it published the most number of articles in Libya, but also

January, 2011, when protests began in Egypt and Syria, followed by Guardian and Peninsula.

New York Times took a lead initially in

In terms of the total number of articles published by four newspapers, Syria is the most intensely covered country, followed by Tunisia and Egypt. Peninsula has the most coverage, where it takes the lead in 12 countries related to the Arab Spring.

Results

0

25

50

2010 12/01

Libya

2011 01/27

2011 08/14

Aug, 20th 2011 Battle of Tripoli

Jan, 14th 2011 Ben Ali toppled Jan, 25th 2011 Mass protest in Egypt Jan, 26th 2011 Protest began in Syria

2011 11/23

NUMBER OF ARAB SPRING ARTICLES PER DAY

< 1000 articles

1000 to 2000 articles

2000 to 4000 articles

> 4000 articles

NUMBER OF ARTICLES PUBLISHED ABOUT EACH COUNTRY

Algeria

We focused on articles published by four newspapers, New York Times, Guardian, Peninsula, and Gulf Times about Arab Spring events. We designed one computer program to extract articles from the websites of these newspapers and a second program to analyze the content of these articles to determine which were related to the Arab Spring. We used these results to contrast the coverage by the four newspapers.

Methodology

Mauritania

Morocco

Tunisia

2012 02/06

Oct, 23th 2011 Tunisians vote Nov, 23th 2011 Saleh resigns Jun, 2nd 2011 Life sentence for Mubarak

Sudan

Egypt

Palestine

Lebanon

2012 10/30

May - Oct 2012 Syrian Massacres

Jordan

Syria

2012 12/22

Saudi

0

35

70

10

20

2010 12/01

2011 03/01

U.A.E.

Notable Difficulties

Qingjia Jiang Northwestern University qingjiajiang2016@u.northwestern.edu

Peninsula: 34,027

2011 09/01

Gulf-Times: 83,877

2011 11/01

articles downloaded articles downloaded

2012 10/01

2012 12/22

Libya

Syria

David Emmanuel Gray, Ph.D.

Faculty Advisor:

Gulf Times were only willing to provide us with printed copies of their articles, but then they were always unavailable for us to collect them from Oman their offices. Meanwhile, Peninsula did give us digital copies of articles they themselves considered Arab Spring related but then they declined to respond when asked for more. In response, we wrote a new computer program to collect all their articles in digital format from their respective websites.

Bahrain

Qatar

Yemen

Kuwait

Arab Spring uprisings across Middle Eastern and North African nations have attracted substantial media attention worldwide. In this study, we explore whether there is any notable variation concerning how different newspapers respond to the major events of these uprisings.

Arab Spring Newspaper Coverage

Jiyda Mint Moussa Carnegie Mellon University jiyda@cmu.edu


Let the Grades Flow! Authors Syed T. Haider (BA 2014) Aveed A. Sheikh (BA 2014)

Faculty Advisor David Emmanuel Gray, Ph.D.

Category General Education

Abstract

Csikszentmihalyi refers to flow, or “optimal experience,” as a state of complete immersion in meaningful activity. A person in flow is characterized by feelings of enjoyment or satisfaction in the ongoing process of that activity. Csikszentmihalyi’s theory entails that an individual is considered in flow when the activity’s perceived difficulty balances with that person’s perceived skills. In order to determine the implications of this theory on students, we explored the flow phenomenon inside a classroom environment. This involved focusing on a particular course for the entirety of the semester, and gauging how student skill and challenges compared and its implications on their academic performance in the class. The main question we endeavor to answer: Are students in flow better academic performers as compared to the students who are not in flow? The samples for this research were students enrolled in the course “Introduction to Logical Reasoning”, offered jointly by Carnegie Mellon University in Qatar and Northwestern University in Qatar. After every class lecture, quiz, workshop, and exam, students were asked to fill out a one-page survey. Each survey response was then assigned to one of the following four groups: 1. Apathetic students showed a lower than average score on challenge and skills, 2. Anxious students showed a higher than average score on challenge and lower than average score on skill, 3. Bored students showed a higher than average score level on skill, and a lower than average score level on challenge, and 4. Flow students showed higher than average score on both challenge and skills. An end of semester survey also had students compare their current abilities in logic to what they started with. The results indicated a difference in the grades and perceived learning amongst the four groups. Anxious students reported the highest perceived learning distance among the four groups of students. Moreover, anxious students showed a positive correlation with respect to perceived learning distances, whereas the other three groups showed a negative correlation, with bored students reporting the highest. The research findings even suggest that student engagement level in terms of being apathetic, bored, and anxious or in flow affects their class performance, with the best performance being observed from students who are in flow. Even though some students showed signs of being apathetic, still they seemed to enjoy the material. Similarly, apathetic students showed more interest as compared to anxious students. We conclude that having a structured classroom environment that initiates flow can help students perform the best in class; as they are engaged in the experience, which leads to a better understanding of the material and higher grades.

43


Let the Grades Flow! Aveed Sheikh

aasheikh@qatar.cmu.edu

Syed Tanveer Haider sthaider@qatar.cmu.edu

Faculty Advisor: David Emmanuel Gray, Ph.D 1 - Flow

2 - Assumptions

3 - Hypothesis

4 - Approach

Csikszentmihalyi refers to flow, or “optimal experience”, as a state of complete immersion in meaningful activity. According to this theory, an individual is considered in flow when the activity’s perceived difficulty balances with that person’s perceived skills.

We predict that students, on average, perform better at academics when they are in flow.

5 - Results 0.00 - 0.99: R; 1.00 - 1.99: D; 2.00 - 2.99: C; 3.00 - 3.99: B; 4.00 - 4.99: A

In this research study, we are analyzing whether students in flow perform better in class as reflected through their grades and whether these students show a greater learning distance than anxious, apathetic or bored students.

After every class lecture, quiz, workshop, and exam, CMU-Q & NU-Q students enrolled in the course “Introduction to Logical Reasoning” were asked to fill out a one-page survey. Each survey response was then assigned to the four groups after calculating the relative z - scores – Flow, Anxious, Apathetic, or Boredom.

Flow

Apathy

Challenge

Anxiety

Boredom Skill

Correlation Coefficient

Flow

Anxiety

Boredom

Apathy

Learning Distance

-0.016

0.33

-0.31

-0.085

* A larger mean count represents more engagement. (Mean count ranges from 0-4)

6 - Conclusions

Bored students report the highest average grades followed by students who experience flow. The lowest grades were reported by the anxious students. Bored students and those in flow performed the best within all learning channels. Finally, only anxious students showed a positive correlation with perceived learning distances.

8 - References

Csikszentmihalyi, M. Flow: The Psychology of Optimal Experience. New York: Harper and Row.

7 - Further Work

In the future, we hope to extend this research into analyzing similar results in other courses offered at CMU-Q, NU-Q or other universities. The results from the research study also suggest that we need to build a platform that can track and provide immediate feedback to the course instructor on how students are performing in a particular course. Such a platform will help him/her better design the course and structure the classroom in a manner that initiates flow in students, whereby enhancing their learning experience and overall performance.


Decentralized Execution of Multiset Rewriting Rules for Ensembles Author Edmund Lam, Ph.D.

Faculty Advisor Iliano Cervesato, Ph.D.

Category Postgraduate

Abstract Parallel and distributed programming is widely known to be a notoriously difficult endeavor and the search for more effective parallel and distributed programming methodologies rages on. In this presentation, we explore exploiting the concurrent and declarative nature of multiset rewriting as a high-level programming model to express complex synchronization behaviors among an asynchronous â&#x20AC;&#x153;soupâ&#x20AC;? of abstract computation units that we call an ensemble. Specifically, we introduce CHR^e, a concurrent committed-choice rule-based language based on core principles of constraint multiset rewriting. We highlight the technical challenges involved in developing an effective decentralized execution of CHR^e, and briefly highlight our operational semantics of CHR^e that guarantees completeness of distributed multiset rewritings while computing rule matches incrementally. While this project is at a theoretical and prototyping phase, we discuss the future works that will take our work to the domain of applied research and development.

45


Decentralized Execution of Multiset Rewriting Rules for Ensembles Edmund S. L. Lam and Iliano Cervesato

Carnegie Mellon University, Qatar

1. Challenges of Parallel and Distributed Programming 

A notoriously laborious and difficult endeavor

Wide range of technical difficulties (e.g. deadlock, atomicity, fault-tolerance).  Traditional computational problems (e.g. correctness, completeness, termination).  While ensuring scalability and performance effectiveness. 



5. Example: Distributed Hyper-Quicksort Distributed Hyper-Quicksort: Assumes loosely coupled ensembles (network, message passing interface, etc..) - - “Local” sorting algorithm Parallel merge sort rules ... - - Distributed Hyper quicksort rules [X ]sorted(Xs), [X ]leader()\[X ]leaderLinks(G) ⇐⇒ len(G) > 1 | let LG, GG=split(G). [X ]leaderLinks(LG), [head(GG)]leader(), [head(GG)]leaderLinks(GG), {[Y ]median(Xs[len(Xs)/2]) | Y in G} {[Y ]partnerLink (Z ) | Y , Z in zip(LG, GG)} [X ]median(M), [X ]sorted(Xs) ⇐⇒ let Ls, Gs=partition(Xs, M).[X ]leqM(Ls), [X ]grM(Gs) [X ]partnerLink (Y ), [X ]grM(Xs), [Y ]leqM(Ys) ⇐⇒ [X ]leqM(Ys), [Y ]grM(Xs) [X ]leqM(Ls1), [X ]leqM(Ls2) ⇐⇒ [X ]sorted(merge(Ls1, Ls2)) [X ]grM(Gs1), [X ]grM(Gs2) ⇐⇒ [X ]sorted(merge(Gs1, Gs2))

Open research problem:

Distributed programming frameworks (e.g. Map reduce [DG08], Graph Lab [LGK+10], Pregel [MAB+10], Mizan [KKAJ10])  Distributed programming languages (e.g. Erlang [AV90], X10 [SSvP07], NetLog [GW10], Meld [CARG+12])  High-level programming abstractions (e.g. Join Patterns [TR11], Parallel CHR [LS11]) 

We seek an approach that is declarative, based on logical foundations, expressive and concise.  Motivated by chemical reaction equations: 

 

6CO2 + 6H2O → C6H12O6 + 6O2

6. Main Challenges

2. Introducing Rule-Based Multiset Rewriting 



Constraint Handling Rules (CHR) [Fru98] ¨

Rule-based constraint logic programming language.  Based on multiset rewriting over first order predicate terms, called CHR constraints.  Concurrent, committed choice and declarative.

CHR programs consist of a set of CHR rules of the following form: 



Incremental matching Termination on quiescence  Interrupt (event) driven matching 





Informally means: If we have P and S such that G is satisfiable, replace S with B. base : gcd(0) ⇐⇒ true reduce : gcd(N) \ gcd(M) ⇐⇒ 0 < N ∧ N ≤ M | gcd(M-N)      

gcd(9), gcd(6), gcd(3) gcd(3), gcd(6), gcd(3) gcd(3), gcd(3), gcd(3) gcd(0), gcd(3), gcd(3) gcd(3), gcd(3) gcd(0), gcd(3) gcd(3)

reduce : gcd(6)\gcd(9) ⇐⇒ 0 reduce : gcd(3)\gcd(6) ⇐⇒ 0 reduce : gcd(3)\gcd(3) ⇐⇒ 0 base : gcd(0) ⇐⇒ true reduce : gcd(3)\gcd(3) ⇐⇒ 0 base : gcd(0) ⇐⇒ true

< 6 ∧ 6 ≤ 9 | gcd(3) < 3 ∧ 6 ≤ 9 | gcd(3) < 3 ∧ 6 ≤ 9 | gcd(0) < 3 ∧ 6 ≤ 9 | gcd(0)











Elements are distributed across distinct locations (k1, k2, etc..), each possessing its own multiset of elements. edge(k2 , 1), ..@k1 ←→ edge(k1, 2), edge(k3 , 8), ..@k2  ↓ edge(k1, 10)@k3

Rewrite rules explicitly reference the relative location of constraints:



[l]c specifies that matching c is located at l. Rewrite rules can specify “local” rewriting:

What are the minimal core language features? What extended language features do we need?  What kind of type safety guarantees can we provide?





[X ]unsorted([I]) ⇐⇒ [X ]sorted([I]). [X ]unsorted(Xs) ⇐⇒ len(Xs) > 2 | exists Y . exists Z . let (Ys, Zs) = split(Xs). [Y ]parent(X ), [Y ]unsorted(Ys), [Z ]parent(X ), [Z ]unsorted(Zs). [X ]sorted(Xs), [X ]parent(Y ) ⇐⇒ [Y ]unmerged(Xs). [X ]unmerged(Xs1), [X ]unmerged(Xs2) ⇐⇒ [X ]sorted(merge(Xs1, Xs2))  

New locations “dynamically” created to solve sub-problems. completed sub-problems are transmitted to the “parent” location.

http://www.qatar.cmu.edu/˜sllam/

Fault tolerance and recovery. Serializability of distributed execution.

7. Current Contributions and Results 

Developed an operational semantics for 0-link restricted rewriting Based on CHR refined operational semantics [DSdlBH04]. Decentralized, Incremental, interrupt driven execution.  Proven soundness and completeness (exhaustiveness) of rewriting  



Formalized encoding of n-link restricted rewriting into 0-link restricted rewriting Based on 2 Phase commit n-consensus protocol [ML85]. Optimized encoding for 1-link restricted rewriting  General encoding for n-link restricted rewriting  



Prototype implementation

Implemented in Python, decentralized execution via OpenMPI bindings and thread scheduling via multi-threading libraries.  CHR based optimization of multiset matching (e.g. optimal join ordering, indexing for non-linear patterns, early guard scheduling)  Basic resource mapping: Initial locations mapped to OpenMPI nodes, dynamically created locations mapped to threaded computation at source of creation. 

8. Future Works 

Finalizing language design and high performance implementation C, C++ or Haskell(GHC) as source language Improving high-level feature encodings +  Explore implementation via Pregel [MAB 10] or Mizan [KKAJ10].

[k1]edge(k2 , 1), [k2 ]path(k3 , 8) ⇐⇒ k1!=k3 | [k1]path(k3 , 9)

Parallel mergesort: Assumes tightly coupled ensembles (multicore, shared memory, etc..)

Existing woes and challenges of distributed programming: 

Rewrite rules can specify link-restricted rewriting:

4. Example: Parallel Mergesort

Designing the Language: 

edge(k2, 1), path(k2, 1), path(k2, 10)@k1 ... [k1]path(k2 , 1)\[k1]path(k2 , 10) ⇐⇒ 1 < 10 | true.  edge(k2, 1), path(k2, 1)@k1 ... edge(k2 , 1), ..@k1 ←→ path(k3 , 8), edge(k1, 2), edge(k3 , 8), ..@k2  ↓ edge(k1, 10)@k3  edge(k2 , 1), path(k3 , 9), ..@k1 ←→ path(k3 , 8), edge(k1, 2), edge(k3 , 8), ..@k2  ↓ edge(k1, 10)@k3

Initialization: How are “locations” distributed across actual distributed system? Load-balancing: How are dynamically created “locations” distributed?



base rule : [X ]edge(Y , D)\. ⇐⇒ [X ]path(Y , D). elim rule : [X ]path(Y , D1)\[X ]path(Y , D2) ⇐⇒ D1 < D2 | true. trans rule : [X ]edge(Y , D), [Y ]path(Z , D ) ⇐⇒ X !=Z | [X ]path(Z , D + D ).



Requires that locations X and Y rewrites respective multisets atomicity . In general (n locations involved), its essentially n-consensus problem.

Designing effective mappings from locations to computation resources 

3. CHR e , Distributed Multiset Rewriting for Ensembles 

Execution of link-restricted rewrite rules is non-trivial: [X ]partnerLink (Y ), [X ]grM(Xs), [Y ]leqM(Ys) ⇐⇒ [X ]leqM(Ys), [Y ]grM(Xs)

r : P \ S ⇐⇒ G | B

Example: Greatest common divisor (GCD)

Effective execution of multiset rewriting in decentralized context: 





Data (unsorted numbers) initially distributed across 2n locations. In termination (quiescence), 2n locations are in total order.

 



Improve language design   



Aggregates, linear comprehensions, Datalog style retraction Extending core language New features via encoding in core language

Dealing with unreliable communications and faulty computation resources  

Fault tolerance backends and fault recovery interfaces Improved n-link restriction encodings (via 3 Phase commit [KD95] or Paxos Algorithm [Lam98])

∗ Funded by the Qatar National Research Fund as project NPRP 09-667-1-100 (Effective Programming for Large Distributed Ensembles)

sllam@qatar.cmu.edu


ClusterLoc: Exploiting Short Range Wireless Technologies for Energy Efficient Localization Author Mohammed Tarek Abdellatif

Faculty Advisor Khaled Harras, Ph.D.

Category Postgraduate

Abstract With the ubiquity of WiFi-enabled smartphones, and large-scale access point deployment, WiFi-based localization is one of the most promising indoor localization systems. Existing WiFi localization solutions, however, exhibit high power demand due to the periodic updates required, which raises the barrier for deployment since battery life is a crucial resource. In this poster, we propose a cluster-based localization algorithm, integrate it with our GreenLoc architecture, and evaluate its performance via simulation and prototype implementation. We design and implement a proximity-based clustering algorithm (CLoc) as a representative strategy that we integrate with GreenLoc. CLoc uses low energy wireless technologies, such as Bluetooth, to detect and cluster individuals moving together. It then assigns a group representative to act as a designated cluster-head (CH) that would be constantly tracked. The location of other group members is then inferred so long as they remain within proximity of the corresponding cluster-head. CLoc dynamically handles the merger or splitting of clusters as a result of mobility. We implement a prototype of GreenLoc and test its operation over different Android devices. We also evaluate the impact of our architecture by studying the performance of our proposed CLoc strategy in our prototype as well as via the QualNet simulator to obtain more scalable results. Our initial results show that we can achieve up to 60% energy reduction with a relatively small degradation in localization accuracy averaging 2 meters. This accuracy reduction is non-impactful given the typical applications expected to leverage our system.

47


Exploiting Short Range Wireless Technologies For Energy Efficient Localization M.Tarek Abdellatif, Abderrahmen Mtibaa, and Khaled A. Harras Carnegie Mellon University Qatar

I. LOCALIZATION APPLICATIONS

Indoor Localization

Various technologies Applications: • School campus • Guided tours of museums • Social networking • Hospitals

II. MOTIVATION

Outdoor Localization

Dominated by GPS Applications: • Navigation & Tracking of assets • Social networking • Calling a cab • Emergency road side assistance

Different energy saving strategies [2]

III. SYSTEM ARCHITECTURE

The Cloc clustering algorithm

The GreenLoc architecture

IV. SIMULATION RESULTS

IV. SIMULATION SETUP Random way point

Simulation results using the Qualnet network simulator [1]

Parameters Table

Metrics Table The impact of various joining and splitting confidences on ALE and APC

Group mobility

Scenario Outline

Equations for ADE and APC

The impact of the mobility model and clustering range on localization error and average power consumption [4]

VI. PROTOTYPE IMPLEMENTATION We implement a prototype of GreenLoc [5] that comprises of three parts: (i) a generic localization server Ploc (ii) the GreenLoc architecture implementing CLoc and (iii) GreenLoc mobile client. We test CLoc using 5 android devices; Samsung Galaxy SI, Samsung Galaxy SII, and 3 Google Nexus One phones, all running the Android OS 2.3.

Normalized APC from CLoc when compared with PLoc.

References

The impact of mobility on Ploc, Cloc, and MCLoc

VII. CONCLUDION & FUTURE WORK

Time = 0 mins

Time = 20 mins

• We achieve 60% reduction in localization energy per node with 28% accuracy degradation averaging 2 meters. • Indoor Localization Architecture • We introduced a novel way to track representative nodes using clustering. • Prototype implementation. • Large-scale testing. • Utilizing social data to improve clustering. • Power consumption fairness across nodes.

[1] Qualnet 5.0. http://www.scalable-networks.com. [2] Chuang wen You, Yi-Chao Chen, Ji-Rung Chiang, P. Huang, Hao hua Chu, and Seng-Yong Lau. Sensor-enhanced mobility prediction for energy-efficient localization. In SECON, 2006. [3] Moustafa Youssef and Ashok Agrawala. The Horus WLAN Location Determination System. In ACM Journal of Wireless Networks (WINET), 14:3 June 2007. [4] R. Friedman, A. Kogan, and Y. Krivolapov. On power and throughput tradeoffs of wifi and bluetooth in smartphones. In INFOCOM, 2011. [5] M. Abdellatif, A. Mtibaa, K . Harras, and M. Youssef. GreenLoc: An Energy Efficient Architecture for WiFi-based Indoor Localization on Mobile Phones. In ICC, 2013.


Performance and Cost Analysis of MapReduce Applications on Public Clouds Author Fan Zhang, Ph.D.

Faculty Advisor Majd Sakr, Ph.D.

Category Postgraduate

Abstract: The MapReduce programming model is a widely accepted solution to address the rapid growth of big-data processing demands. Various MapReduce applications with a very large volume of input data can run on an elastic compute cloud composed of many distributed computing instances. A public cloud provider, such as Amazon EC2, offers a spectrum of cloud resources with varying costs. Cloud users typically rent these elastic cloud resources as virtual machines (VMs) in a pay-as-you-go model to have access to large scale cloud resources. However, different applications scale differently based on their type, behavior and effective use of resources available. In this work, we attempt to characterize how MapReduce performance is affected by increased compute resources for a variety of application types. Since resources on public clouds are rented, we carry out a performance cost analysis in order to assess the efficiency of a suite of MapReduce applications at utilizing a range of compute resources. These applications span across data- and compute-intensive benchmarks. Through empirical evidence, we observe a wide variation in speedup (5.2X to 36.7X) and cost (3.6X to 9.7X), across the applications when the cluster size is increased to 64 VMs. Map-intensive applications, such as TermVector and Grep, show a higher speedup as we increase the number of VMs without a significant increase in cost. However, reduce-intensive applications such as Sort exhibit limited speedup and hence cost a lot more since more resources are utilized for a longer period. Given this wide variation, we measure the efficiency of applications to utilize compute resources as the number of VMs is scaled from 1 to 64. We observe a negative slope in efficiency as the number of VMs is increased across all applications. At 64 VMs, the application efficiency range is from 57% down to 8%. Some applications, such as Sort, exhibit a steep negative slope in efficiency when the number of VMs is increased from 2 to 4. WordCount maintains a high efficiency at 4 VMs but exhibits a steep negative slope when increasing the VMs to 8 onwards. Grep on the other hand exhibits a slight but steady negative slope from 2 to 64 VMs. The efficiency of an application can guide cloud users in choosing appropriate computing resources based on compute resource budgets and deadlines. 49


Characterization of MapReduce Applications on Private and Public Cloud Platforms Fan Zhang

Majd Sakr

School of Computer Science, Carnegie Mellon University in Qatar {fanzhan1, msakr}@qatar.cmu.edu

MapReduce Application Types

Hadoop MapReduce Framework

Map-intensive Application: WordCount

Map

Input Data

Local I/O

Reduce

Shuffled Data

Network

Output Data

Network

Reduce-intensive Application: Sort

Input Data

Map

Local I/O

Shuffled Data

Reduce

Network

Output Data Network

Characterization on a Private Cloud Platform

vCPU

Timeline Analysis

Private Cloud Platform

Memory

Characterization on a Public Cloud Platform

Performance

Cost Performance

Public Cloud Platform Cost

Scalability Efficiency

Conclusion

Future Work

1. Performance of Map-intensive applications is highly dependent on the capacity of the compute resources provisioned. Increasing the number of Map tasks will lead to improved performance if sufficient resources exist to enable concurrency. 2. Performance of Reduce-intensive applications is highly dependent on the capacity of the network substrate to shuffle intermediate data in the Reduce phase. 3. A wide variation is observed in speedup and cost across applications when provisioning larger clusters â&#x20AC;&#x201C; 4X speedup at 10X cost to 36X speedup at 5X cost. We attribute this variation to the efficiency of applications to utilize cloud resources.

1. Build a compute cluster with varied bandwidth to further verify the networkdemanding characteristics of Reduce-intensive applications. 2. Develop a viable performance model, which takes into account the Map- and Reduce-intensive phases, and incorporates the impact on performance and cost. 3. Employ a larger cluster on both private and public cloud platforms to evaluate more diverse benchmark applications across a variety of input data sets.


The QALB Project: Building Resources and Systems for the Automatic Correction of Arabic Text Authors Wajdi Zaghouani, Ph.D.

Faculty Advisor Behrang Mohit, Ph.D.

Category Postgraduate

Abstract We present our efforts in building QALB, The Qatar Arabic Language Bank of edits. The first goal of this project is to create a large corpus (approximately 2 million words) of human-corrected Arabic text produced by native speakers, non-native speakers, and machines. The annotated corpus will provide training data for future statistical Arabic errors correction tools. In order to build the QALB corpus, we created comprehensive Arabic errors annotation guidelines to be used by the team of annotators and volunteers participating in this project. Furthermore, we created a web-based annotation tool and an annotation workflow management interface to be used during this project. We will demonstrate the current annotation framework and the outcome of the pilot annotation phase of the project. Once the QALB corpus is ready, we plan to build an automatic Arabic errors correction framework. This project is supported by NPRP grant 4-1058-1-168 from Qatar National Research Fund (QNRF).

51


The QALB Project:

Building Resources and Systems for the Automatic Correction of Arabic Text Wajdi Zaghouani

Behrang Mohit

Ossama Obeid

wajdiz@cmu.edu

oobeid@cmu.edu

1. Problem Statement

How can we fix automatically the errors found in Arabic text ? How can we improve the English-Arabic machine translation output ?

3. The QALB (Qatar Arabic Language Bank ) Project The project is composed of two components:  Corpus of Arabic error corrections.  Automatic correction tool for Arabic.

behrang@cmu.edu

Kemal Oflazer ko@cs.cmu.edu

2. Solution

Build resources and systems for the automatic correction of various Arabic text.

4. QALB Corpus    

Machine Translation output. Native speaker errors (eg. dialectal words). Non-native speaker errors. Student essays

5. QALB ANNOTATION Guidelines

Requirement : comprehensive and coherent annotation guidelines which cover:  Spelling Errors  Punctuation Errors  Lexical Errors  Morphology Errors  Syntactic Errors  Dialectal Usage Correction

Figure 3: The QALB annotation screen.

Figure 4: The QALB annotation pipeline.

7. Ongoing Progress

6. Annotation Tool

 Intuitive interfaces for: • Correcting errors. • Move, merge, split, insert words • Managing annotation workflow.  Automate simple corrections.

 Annotation  Automatic Correction Tool

8. Summer job opportunity  Contributing to a research effort on improving Arabic digital content.

Figure 6: Word correction functions. Figure 5: Split word correction example.

Acknowledgement

This research was supported by Qatar National Research Fund (QNRF), NPRP grant 4-1058-1-168

 Requirement: high school level Arabic knowledge.


Type-Based Productivity of Stream Definitions Author Jorge Luis Sacchini, Ph.D.

Category Postgraduate

Abstract Infinite data structures are used for defining non-terminating processes such as webservers, operating systems, or network protocols. A prime ex- ample of an infinite data structure is a stream, which represents an infinite sequence of elements. A desirable property of a stream definition is productivity. A stream definition is productive if computing any of its elements takes a finite amount of time. In other words, productive definitions cannot get stuck. Proof assistants and theorem provers enforce productivity of stream definitions in order to ensure logical consistency. However, they usually rely on syntactic methods that are relatively restrictive and difficult to use. In this work we propose a type-based method for ensuring productivity of stream definitions, which is more expressive and intuitive than the syntactic-based methods normally used. Furthermore, type-based methods are compositional, which eases the task of developing large specifications. We extended the underlying theory of the Coq proof assistant with a type-based productivity checker and showed that the approach preserves soundness (i.e. logical consistency).

53


• • •

• •



server :: Requests → Answers server(ClientReq : reqs) = ServerAns : server reqs •

server :: Requests → Answers server(ClientReq : reqs) = server reqs

server :: Requests → Answers server(CReq : reqs) = if (overloaded) then server reqs else SrvAns : server reqs

• •












For more than a century, Carnegie Mellon University has been inspiring innovations that change the world. Consistently top ranked, Carnegie Mellon has more than 12,000 students, 90,000 alumni and 5,000 faculty and staff globally. In 2004, Qatar Foundation invited Carnegie Mellon to join Education City, a groundbreaking center for scholarship and research. Students from more than 40 different countries enroll at our world-class facilities in Education City. Carnegie Mellon Qatar offers undergraduate programs in biological sciences, business administration, computational biology, computer science and information systems. Carnegie Mellon is firmly committed to Qatarâ&#x20AC;&#x2122;s National Vision 2030 by developing people, society, the economy and the environment. Learn more at www.qatar.cmu.edu


P. O . B o x 2 4 8 6 6 | E d u c a t i o n C i t y, D o h a , Q a t a r | P h : + 9 7 4 4 4 5 4 8 4 0 0 | F a x : + 9 7 4 4 4 5 4 8 4 1 0 | w w w. q a t a r. c m u . e d u


Meeting of the Minds, 2013