Page 1

Project Acronym: BRIM project Version: 1.0 Contact: Lee McCluskey Date: 30/09/10

JISC Final Report Project Information

Project Acronym


Project Title

Using Business Process Management Tools and Methods for Building Research Information Management

Start Date

1 March 2010

Lead Institution

University of Huddersfield

Project Director

Dr Liz Towns-Andrews

Project Manager & contact details

Professor Lee McCluskey Director of Research, Computing and Engineering Telephone: +44 (0) 1484 472247

Partner Institutions


Project Web URL

Programme Name (and number)

Information Environment Programme: Research Management Information 11/09

Programme Manager

Neil Jacobs

30 September 2010

End Date

Document Name

JISC Final Report

Document Title Reporting Period Author(s) & project role

Kirsty Taylor (Business Intelligence and Marketing Manager) Project Officer




JISC Final Report

URL ď Ż Project and JISC internal


ď Ż General dissemination

Document History Version 1.0

Date 30.09.10

Page 1 of 35 Document title: JISC Final Report Last updated: April 2007


Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Table of Contents JISC Final Report ....................................................................................................... 1 Table of Contents ....................................................................................................... 2 Acknowledgements .................................................................................................... 3 Executive Summary ................................................................................................... 4 Background ................................................................................................................ 5 Aims and Objectives................................................................................................... 6 Methodology ............................................................................................................... 7 Implementation ........................................................................................................... 9 Outputs and Results ................................................................................................. 16 Outcomes ................................................................................................................. 20 Conclusions .............................................................................................................. 21 Implications .............................................................................................................. 22 References ............................................................................................................... 23 Appendix A - ePrOnto: OWL-Based Ontology for ePrints Information Management 24 APPENDIX B - University business computer systems and their functions.............. 32 APPENDIX C – RIMS Overview ............................................................................... 34

Page 2 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Acknowledgements The project ‘Using Business Process Management Tools and Methods for Building Research Information Management (BRIM)’ was funded as part of the Information Environment Programme through JISC. We would like to gratefully acknowledge the following individuals for their contribution to the project and their ongoing support: Name

Job Title

Professor Andrew Ball

Pro Vice-Chancellor (Research & Enterprise)

Dr Liz Towns-Andrews

Director of Research and Enterprise

Professor Lee McCluskey

Director of Research, Computing and Engineering

Professor John Lancaster

Director of Computing and Library Services

Jimmy Dane

Business Application Manager

Sam Flanagan

Information Systems Developer

Annabel Holland

Head of Research and Graduate Education

Wayne Keating

School Finance Manager (Human & Health Sciences)

Kate Mitchell

CRM Manager

Denise Ogden

Manager – Bids and Contracts

Joanna Olszewska

Research Assistant, Computing and Engineering

Krishnananda Pilicudale

Business Application Manager

Ian Pitchford

PGR and REF Manager

Alan Radley

Head of Computing Services

Ron Simpson

Contract Consultant

Graham Stone

Repository Manager

Kirsty Taylor

Business Intelligence and Marketing Manager

Yvonne Whiting

System Support & Development Manager (Finance)

Page 3 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Executive Summary Universities have large computing infrastructures mainly concerned with delivering academic services to their students and staff. They also require computing systems to support their core business functions such as finance, purchasing, HR, and, the subject of this project – research information. This project was aimed at investigating the feasibility of implementing research information management (RIM) tools to satisfy the management needs of a University which is growing its research activity significantly. The University of Huddersfield planned to utilise the state of the art information processing ideas such as business process modelling, service orientation, and knowledge bases called ontologies, and integrate them with legacy core business computing systems mentioned above. The idea was to investigate a method to expose existing business systems as services usable by other computing processes, in such a way that their interfaces were self-describing and conforming to an international standard. The project started with interviews with RIM stakeholders (School research administrators, research senior management, finance staff and research budget holders) in order to elicit requirements. Current RIM data, and current University core business systems were assessed. The approach taken was to split the project into 3 tasks: the main task, discussed in the body of this report, was to capture RIM requirements and create a prototype system that integrated with existing business processes using standard interfaces. This was supported by two other tasks: one was to evaluate the existing core business process tools with respect to their capabilities to share data and processes. The other was to investigate the use of rich data representations called ontologies for representing and storing research information, in particular publication information. To make the aims attainable in a 6 month period we focussed on two areas of RIM: research project bidding process, and the research publication data repository. The initial findings indicated that it would be difficult to integrate and use data from existing computer systems in the University in a fully standardised, service oriented approach, but it would be possible, nevertheless, to create a research information layer which integrated with and harvested data from existing systems. To demonstrate this, during the project we developed such a pilot system, using technology that integrated data from existing computer systems in a standard, component-based way, and stored it in a model based on a developing standard for RIM called CERIF. The technology allowed the systems to be viewed as “services”, but was reliant on the systems being produced by a particular software company. Hence, although we had to compromise on the “openness” of the technology, the solution adopted preserves features of the service-oriented approach and utilises the CERIF standard. A key result of the project was that it demonstrated how data stored in a University publications repository can be reconciled with other University computer systems to create accurate management information about staff publications.

Page 4 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Background As a University we face pressing demands to develop sophisticated information systems software to support our business processes, including those for research information management (RIM). Off the shelf software lacks flexibility and does not seem to meet the requirements of a RIM system and yet, bespoke software is usually prohibitively expensive to develop and maintain. It is has become apparent through the University’s planning process that there is an emerging need to take a strategic approach to RIM, focusing on horizontal business processes. The University currently stores RIM information on numerous disconnected systems. These separate systems support vertical functions of the business rather than horizontal processes. These systems are subject to largely independent governance and cross functional interoperability is not currently within the remit of systems managers/directors. This presents problems for senior management when they wish to see a unified view of processes for strategic decision making and reporting. There is now a strong commitment from the Senior Management Team to investigate the ways in which these systems can be used to improve our business and therefore deliver better value to our customers.

Page 5 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Aims and Objectives The aims of the project, as laid out in our initial project plan, were to investigate the representation of research information management (RIM) data in an open, accessible form, to investigate business processes surrounding RIM, and hence investigate supporting the task of RIM at Huddersfield with business process support tools. The benefits lie in extracting and uniting enterprise research information, making it a horizontal resource, and in assisting RIM in the University with business process management tools. To this end, we set out:     

to identify and document stakeholders and their interests to create detailed process models of RIM business processes to analyse processes and identify possible improvements to identify and document capabilities of existing core systems with respect to Service Oriented Architecture (SOA) [1], and the potential to implement an SOA-based system to share our knowledge and project outputs with JISC and the HE community

The overarching objective of RIM is to foster and monitor research work carried out throughout the University. Research work includes commercial projects, publically funded projects, research publications and work by staff towards higher degrees. The university currently has a number of central systems in place to assist in this work. There are financial systems to manage funded projects both commercial and publically funded. There are “Human Resource” systems to manage and monitor staff working towards higher degrees and there is a repository system to help staff publicise and make available their research publications. The project did not change in objectives, or outcomes. After initial analysis at the beginning of the project, however, the way we achieved the objectives, and hence some of the outputs, changed. The reasons were: 1. Some key RIM data was not being stored in core University computer systems 2. The initial orientation of the project was to attempt to use general BPM methods. For BPM systems to work effectively, a high degree of open architecture, open communications protocols and systems interoperability is required. We could find little evidence that these core systems could be used within an open architecture as originally planned. The project’s objectives were met using a proprietary package that interfaced with all of the existing core services, as will be detailed below. Additionally, the proof of concept implementation changed. Originally we had planned to produce an ontology tool. In the project, we decided to concentrate on solving the “identity” problem for publications – that is making the mapping between staff (as identified by their staff id) to publication author more accurate.

Page 6 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Methodology The initial approach taken was to  interview RIM stakeholders in order to elicit requirements  assess current RIM data, its accessibility, accuracy etc  assess current University core systems  investigate Business Process Modelling tools, and their applicability in the context of the University’s current core systems.

In starting the project the project team tried to identify key stakeholders within the system and interview them with a view to both establish current practice, and known weaknesses in the current system. In the event we conducted ten interviews, and where appropriate these were recorded in a set of UML “Use Case” diagrams. The diagrams provide overview descriptions of the activities undertaken and further identify the key stakeholders involved. In some cases where the complexity of the activity justified the approach “BPMN” (Business Process Modelling Notation) diagrams were also produced. In most cases these diagrams simply reflected current practice but in others they reflected what the stakeholders believed should happen rather than what currently happened. Whereas there are other methods of initial requirements capture, the advantage to the project team in using these relatively simple diagramming techniques was that it allowed the interviewee to quickly verify that the interviewer had correctly understood the activities at an appropriate level of generality. After the initial requirements phase, the project split into 3 parts: 1. The main thrust of the project was to use process modelling and workflow

tools, drawing on data from University core systems horizontally, in order to capture and animate some of the RIM processes, in a manner that supports future enhancements, process interoperability and software extensibility. In particular, the CERIF model [2] was to be used to capture the data. The small proof of concept project planned was allied to this end: during initial analysis of publications data, it became apparent that the information and metrics on staff publications was deficient in two ways (a) the mapping between staff (as identified by their staff id) and publication author was in some cases incorrect (this is an aspect of the perennial identity problem) (b) insufficient data had been collected. The proof of concept application was then aimed at solving these two problems.

Page 7 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

2. University core systems were evaluated with respect to their readiness to fit

into an open architecture. The Open Group Architecture Framework (TOGAF) method was considered for the evaluation, as it represents a well known high level methodology for open systems development [3]. A researcher investigated each one of the University’s legacy systems and attempted to assess it on how "service ready" it was. While this led to the production of the table in appendix B, it also led to the production of a textual report describing the systems in terms of purpose, functionality and specification. The information was obtained through interviews with stakeholders, as well as using system documentation. The amount of information that we could access on each system varied, however, and identifying service oriented features proved very difficult. Therefore within the time scales and man power available to the project team the full adoption of this approach was considered infeasible and a more informal assessment was carried out. 3. The feasibility, usefulness and method of mapping publication information

across to a standard ontology language was evaluated.

Page 8 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Implementation The implementation was carried out in 3 parallel tasks, reflecting the 3 part split detailed above. The findings of the main thrust will be detailed below, with an overview of outcomes and lessons learned from the whole project. The report on developing the ontology is annexed in Appendix A, and an outline user manual of the pilot implementation given in Appendix C. Appendix B lists the main University business computer systems and their functions.

Functional Area Use Cases We provide here key examples of the functional areas studied and samples of the diagrams produced.

F1: Financial Monitoring and Control of Funded Projects The primary central contact between research groups and University management is through the Finance Department. Research groups are largely dependent on central finance to assist in preparing full economic costs (fec) of proposed project bids. They are then reliant on Finance to provide detailed management of funds when bids have been successfully won. An overview of the current bidding process as seen centrally is shown in Figure 1. The primary feature of this process is that from the point of view of central administration the bidding process for a funded project starts with a request by a research group for assistance in producing a costing of the proposed project. From the perspective of research groups a great deal of work will have taken place prior to this point and detailed costing is in many cases only required immediately before submission of the bid. Research groups will often work with a rough costing which they can prepare themselves prior to this point. This practice tends to produce bottlenecks in providing detailed costing as Finance may not be able to respond to requests at the speed desired when a bid is all but ready to submit. The costing as a consequence of the haste is often done in a mechanical manner that denies the opportunity to negotiate on the appropriate model to apply to the individual bid. The private nature of the initial bidding process also tends to mean that other groups with a potential interest in the call for bids may be unaware of the work being done by the first group and hence miss opportunities for potential collaboration or potential conflict. It also denies the central organisation of early oversight of bids and consequently makes it difficult to develop strategies to manage the range of opportunities to be pursued.

Page 9 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 1 Use Case: Bidding Process

An idealised version of the Bidding process is shown in Figure 2. In this idealised process we have both a central database and a central repository for bid documents. An advantage of such a system is that the central database and repository starts being populated with bid details as soon as a call for bids is identified as being relevant to staff within the institution. The central repository becomes a growing electronic store of information which can be made readily available to research groups and central administration. We also see that the requirement to generate costing is in the middle of the process rather than the supposed start. There is also no reason why the calculating of a costing at full economic cost or otherwise cannot be largely automated once the requirements of the bid have been specified, notwithstanding the need to allow some flexibility in the model applied. The process as sketched in the BPMN diagram has the potential both to empower research groups and schools and to provide central administration with higher quality information about upcoming bids. It is recognised that putting a version of the bidding process conforming to the above model into place would be a significant effort for the Institution and would involve both initial and concurrent costs if commercial packages were purchased to meet the perceived need. Commercial packages that are capable of implementing the desired process do exist and are in use at some other Institutions of Higher Education.

Page 10 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 2 BPMN Model: Bidding Process

F2: Planning and Monitoring of Research Use Case During the year central administration carries out a number of exercises where an overview of the totality of research that is carried out in the University has to be prepared. As schools are accountable for the research they do, providing this overview requires gathering statistics about research organised by school. A typical example of this effort is required in preparation for the annual “Planning and Accountability� conference held every year where school representatives explain what has been achieved in the past year and negotiate about targets for the coming year. A use case diagram giving an overview of the work involved in preparing for this conference is shown in Figure 3. On the bottom half of the use case we see that drawing together the information for the conference report document requires data from a number of diverse sources, including a number of the Universities primary information systems. Currently there is an electronic system in place to gather and collate much of this information but the result has suffered from a lack of perceived accuracy in some of the figures provided and has required manual intervention in other areas to prepare the figures for incorporating in the report. Superficially, it would seem that as the report basically requires counting and summing the information that should already be stored in the primary information systems. No manual intervention should be required nor should there be uncertainty in the resulting figures.

Page 11 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 3 Use Case: Planning and Accountability

A deeper analysis of the primary systems reveals in some cases that the required data is not being adequately stored in a form amenable to automated processing. An example of this occurs with the publication repository. The function of the repository is to allow academic staff to make their publications available to as wide an audience as possible. To this end the responsibility of entering data rests with academic staff themselves, though there is some central checking of the quality of data entered. The principal difficulty lies when members of staff enter their co-contributors details.

Page 12 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

This is an error prone process. Many second and third authors are entered by name only and in many of these cases a familiar form of their name is used rather than the form officially recognised by the University. The consequence is that a great deal of manual reconciliation of names has to take place before the information stored in the repository can be used to accurately provide statistics on the publication activities of staff throughout the university. This problem also manifests itself in other university systems, though generally in a less acute manner. There is no centrally enforced uniform way of identifying members of staff that is used across all university systems. Payroll numbers are often used to identify members of staff but not all individuals associated with the university and involved with research are on the payroll and can consequently be identified by such unique codes.

Current Information Systems at The University of Huddersfield A brief overview of the universities current IT systems is shown in Figure 4 with connections showing primary exchanges of data between systems. A brief description of each system is provided in Appendix B. PINS BERT


LCMS External examiners database




Active Directory

Wisdom Sharepoint


HR System

Research office database R&E data warehouse

Research grants and contracts

Page 13 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 4 University Core Systems

The R&E data warehouse is the system we are developing as the pilot system with a view to it becoming the main research information system. The R&E system prior to this project did exist in an embryonic form and did provide some reports for management but they are either replaced by the reports of the pilot system or we envisage will be replaced as the pilot model is expanded in scope in the future.

Existing Systems and SOA Architecture Adopting a SOA is often associated with the identification and mark-up of processes as “web services�. To do this, one uses the Web Service Description Language (WSDL) as a kind of universal API to allow the description of services and their messages regardless of the used message format or network protocols to communicate between applications. WSDL is an XML format that describes the available services as a set of network endpoints that operate using messages containing document-oriented or procedure oriented information. WSDL files could be used with web services to describe each service in detail and define what operations it knows and how to perform them. After exposing the functions provided by the university’s IT resources as web services, WSDL files could be used in the University to facilitate interoperability by bridging disparate middleware component technologies and easing client/server implementation. Our investigation found that the key University systems that a new RIMS system would require to interoperate with do not currently provide an adequate web service interface. The two key systems where this was the case were the eprints publications repository and the HR database. The systems at Huddersfield are primarily third party commercial systems managed at Huddersfield but not developed in-house. Consequently, though web service interfaces could potentially be developed for all the systems, it is not in our hands to do so. However, as our pilot system has demonstrated, it is possible to integrate our various systems at the data storage level using more traditional methods. The potential drawback in this approach is that we are not re-using the processing capabilities of the various systems - we are primarily re-using data. The approach does allow for some reuse of processes, if they are incorporated in accessible stored procedures, though the legacy systems are not designed with this as a primary requirement. The main disadvantage in the approach we have used is the effort required in understanding the existing systems needed to be able to extract data from them. Development work would have to take account of the various data models used in each system: in contrast to this a fully developed SOA system would have a uniform way of describing the data and functionality provided by the systems. Despite the potential advantages of developing such a SOA system across the Universities systems we felt that the commitment and effort required to specify the requirements and have the interfaces developed could not be justified by the potential benefits.

Page 14 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Requirements to more fully develop a CRIS system at Huddersfield would be much more achievable by continuing to use the strategy of using SSIS packages to access data at the database or pseudo database level. However, while not being able to efficiently bring existing systems together into an SOA, we could argue that we could nevertheless make some progress. Assuming the RIM layer as piloted at Huddersfield successfully integrated resources from legacy systems to add value to current data, then one could envisage a range of services being plugged into the RIM layer "from above". In this way, the RIM layer is acting as a service: it could provide "cleaned" research data to a range of possibly internal and external clients in a SOA fashion.

Page 15 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Outputs and Results Contrary to expectation, the review of business processes found that the problem of accuracy of management information was not simply a problem of gathering the data from the front line systems. Adding to that difficulty was the problem that not all the required data was being stored in the computerised systems and that in some cases where it was being stored it was not being stored in a manner that allowed automatic reconciliation with data stored elsewhere in the universities systems. This central finding presents a choice in terms of strategy in trying to overcome the problem. One approach would be to revisit the systems already in place and try and improve their quality in terms of their ability to accurately and uniformly capture the data required. A second approach would be to build a new layer to the overall information management system to both coordinate the information in the existing system but also to provide mechanisms for missing data to be stored and provide custom routines to reconcile the data stored in the various front line systems.

Solution Chosen The decision of the team was a combination of the two strategies outlined above. In the case of the existing system to support bidding for funded projects it was decided to recommend that new systems be put in place and that this could be done by the purchase of commercially available systems already in use within the higher education sector. A new project for the specification and purchase of software and associated consultancy would be responsible for putting such a system in place. Replacing or significantly amending other front line systems within the University was not judged to be practicable hence it was felt that there was still a requirement that a new tier of information system targeted at research needed to be constructed. To facilitate this the team decided to focus the pilot system envisaged in the project bid on determining the architecture of the new information layer and showing the viability of the architecture by implementing a segment of the final system. This implementation is referred to as the R&E data warehouse in the diagram of University systems above.

Architecture of the Pilot System The pilot system envisages a software layer providing the Research Information Management System sitting above the front line university systems as shown in Figure 5.

Page 16 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10







Figure 5 Pilot Architecture

How to create the new information layer was a major challenge in the project. We considered two alternatives. One that the new layer be largely virtual in the sense that little or no data be stored in the layer. In this case it will consist of software that draws its data from the existing lower layer systems in an “as needed” manner. Ideally this would be facilitated by each of the lower layered systems exposing services, perhaps as web services, that enabled data to be extracted from the base system. Though desirable the team felt that this approach was not achievable for the following reasons: 1. the lack of uniformity in the way services were exposed by the different front line systems 2. a virtual system would require that considerable work be done to reconcile information between it and the base system; this would needlessly have to be repeated whenever the ‘RIMS’ system was used 3. the opportunity to store value added information additional to the information already being stored would be lost. The decision was then made to make the ‘RIMS’ layer a data storage layer in its own right. Where data storage was duplicated, the master copy would always reside in the base system, and the top layer would always be updated when changes were made to the base systems.

CERIF The Common European Research Information Format ‘CERIF’ provides a reference data model for research information. This model addresses many of the problems of storing information about research typically required by management and by external bodies, hence it appeared appropriate to use this as a starting point in creating the ‘RIMS’ layer at Huddersfield. The pilot implementation of the ‘RIMS’ layer, it was decided, should only focus on a portion of the CERIF model.

Page 17 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

The decision to rework and buy in a system for the BIDS process meant that the portion of the model relating to projects could not yet be tackled. One of the major features of the CERIF model relates to storing information in multiple languages but at Huddersfield there is almost no need for this, consequently implementing the multi language features simply add complexity for no significant benefit and have not been implemented. Hence, in implementing the pilot we needed to decide on how closely to adhere to the published CERIF model. At this stage, given that external bodies are not yet requesting data in the CERIF Interchange format, we felt that the internal information needs of Huddersfield should be considered primary. This decision had two consequences: first it meant that we did not feel it necessary to extract from our base systems all of the data envisaged to be stored in the reference model, some of it was superfluous to internal needs. Secondly, we felt free to add data and fields to meet current needs at Huddersfield. To facilitate future requirements we adopted a rigorous naming convention to record where we were storing data that matched the CERIF model and where we were adding local data. This decision should allow us in the future if required to easily map the information we store to the CERIF interchange format. It would of course still leave us with difficult problems of gathering the data where we have omitted to implement some of the fields, having currently no local need for that data. The primary benefit we were reaping was that by adopting much of the structure of CERIF we were being guided in understanding and capturing the implicit structure of the data we are required to manipulate.

Scope of the Pilot RIMS The pilot ‘RIMS’ system is limited to providing information on staff research publications (refer to the description of the Pilot system in appendix C). In doing this it draws its primary data from the ePrints repository. In focusing on this area it provided the opportunity of resolving the considerable problems of identity present in the current ePrints data and it allowed the opportunity to add value to that data by storing additional information about publications. To resolve the problems of identity of authors and contributors we linked the data extracted from ePrints to data in ‘Active Directory’ (AD) which stores staff email addresses, names, schools to which they belong. We also link to the Human Resources data base ‘HR’ to access unique identifiers for staff along with names and other personal data. The links do not allow us to automatically resolve all issues of identity but where they cannot be so resolved we have built into the human interface to ‘RIMS’ methods to alert staff of where there are problems and to provide additional information to resolve the outstanding information. We have also provided an interface to allow individual members of staff to check the publications data being held against them and to allow them to provide additional information that we anticipate will be required for future REF exercises. The manual checking and adding of information also allows changes to be fed back into ePrints and other core systems, where inconsistencies are found, with the highly desirable consequence of improving the quality of data visible there.

Use of Business Process Tools The project’s original project plan and planned outputs were adjusted somewhat during the project as the implications of the diverse legacy systems at Huddersfield were understood. The use of modern BPM tools such as Intalio Designer largely presuppose that legacy systems either expose well defined web service interfaces or

Page 18 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

that such interfaces can easily be created. Had this assumption been true along with the assumption that all the required data is already stored in the various systems, the methods and tools would have been a good fit and we could have built the ‘virtual’ RIMS system described above. As neither of these assumptions was met, we had to adopt a slightly different approach to systems’ integration. Huddersfield’s legacy systems are all hosted on Microsoft platforms and are generally Microsoft compatible. Accordingly, we were able to use Microsoft’s Business Intelligence platform to perform the integration called SSIS. The SSIS packages we were able to build are similar to other BPM tools in that they provide a largely graphical design for the business logic, but it provides a range of programmatic methods to link up to and extract data from databases and structured files such as excel spreadsheet files. SSIS allowed us to operate with systems that provided no web service interface. Our experience was that the Business Intelligence suit provided us with impressive capabilities to integrate legacy systems but would not necessarily do so had we not been working on Microsoft platforms exclusively. The time and effort involved in implementing SQL SIS is difficult to measure as the SSIS package development went hand in hand with developing both the miniCRIS database and with learning about the structure of sources of raw data for the mini VRIS such as the ePrints, MtSQL database and the implementation of AD. However, with a very clear spec and an expert in developing SSIS packages it is felt that this could be achived in around one week, although at Huddersfield it took around three to four weeks of work.

Page 19 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Outcomes Our aim of investigating data representation and business process within research information management (RIM) data in an open, accessible form has been has been carried out and produced several outcomes, as follows.  We have elicited and analysed a range of business processes surrounding RIM detailed above, and, focusing on particular areas, shown the potential of open, horizontal process by demonstrating their application to identity resolution for in-house publication authors. We have shown how the CERIF model can we used as a basis for a RIM layer which draws its data from supporting layers.  We have analysed existing core business processes and found them relatively closed, with little capability to enable them to fit into an open service architecture. To a certain extent, given (a) the security issues of the data (b) the buy-in to one particular software company, this is not surprising. However, through the pilot implementation we have shown the capability of a proprietary software tool SSIS for integrating processes in a BPM fashion.  We have contributed to the solution of the problem described by Day in [4]: to reconcile the divergence of research management needs, and the requirements of publication repositories, by spanning systems with an extra RIM process layer.  We have looked to the future in data representation and conducted a pilot study in the creation of an ontology for representing and storing publications data. This strand of the project is detailed in Appendix A.

Page 20 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Conclusions The benefits and feasibility of designing a RIM system as a horizontal layer across the core University business processes has been demonstrated, though the use of open architectures and BPM tools that rested on them were replaced by necessity with a proprietary workflow tool. The method of eliciting process workflow using BPM notations for RIM activities has been shown to be useful, and in the process we identified data that was needed, but was not being collected (eg publication annotations), and data that was inaccurate (eg publication identity). Requirements to more fully develop a CRIS system at Huddersfield would be much  more achievable by continuing to use the strategy of using SSIS packages to  access data at the database or pseudo database level. However, while not being  able to efficiently bring existing systems together into an SOA, we could  argue that we could nevertheless make some progress. Assuming the RIM layer as  piloted at Huddersfield successfully integrated resources from legacy systems  to add value to current data, then one could envisage a range of services  being plugged into the RIM layer "from above". In this way, the RIM layer is  acting as a service: it could provide "cleaned" research data to a range of  possibly internal and external clients in a SOA fashion. 

The project’s main innovation was to use a version of the CERIF data design as the basis for a RIM layer which lay horizontally across a set of core University business processes. A prototype implementation has been created and evaluated on this basis. A more detailed description is given in appendix C. While some features of CERIF (eg multilingual fields) were not found relevant, the overall structure was found adequate for the functions of the layer that we prototyped.

Page 21 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Implications We attempted a combination of the two strategies for RIM: revisit the systems already in place and try and improve their quality in terms of their ability to accurately and uniformly capture the data required; and to build a new RIM layer over the existing information systems. While time did not permit us to evaluate the resulting pilot system rigorously, we did manage to coordinate the information in the existing systems and to provide mechanisms for missing data to be stored and provide custom routines to reconcile the data stored in the various front line systems. Hence we recommend that, within an University environment with similar systems to ours, that an enhanced RIM could be constructed as a horizontal layer interacting with existing systems, in particular the publications repository. An obvious development to the pilot system is to extend it so that it can automatically generate, for example, all typical Key Performance Indicators (KPIs) required by University research managers, such as publications, grants, research degrees, esteem factors of staff. The feasibility and architectural concepts have been put in place for that during this project: only development work would be required to complete the system’s functions. An enhancement to the full development of the RIM layer would be to use it to channel publication data to an ontology framework (such as that set up in ePronto in Appendix A). This would give the opportunity of providing richer management information and analysis by utilising the additional power of the ontology query languages.

Page 22 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

References [1] “Briefing paper on the e-Framework and Service-Oriented Architecture” JISC (2006). Available at: [last reviewed 15/9/2010] [2] [3] “Applying the ADM at Different Enterprise Levels”. The open group (2009). Available at: [last reviewed 15/9/2010] [4] "Models for integrating institutional repositories and research information management systems", Michael Day, Proceedings of the Workshop on CRIS, CERIF and Institutional Repositories, CNR, Rome, 2010.

Page 23 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Appendix A - ePrOnto: OWL-Based Ontology for ePrints Information Management Joanna Isabelle Olszewska School of Computing and Engineering, University of Huddersfield, Queensgate, Huddersfield, HD1 3DH, United Kingdom {}

1. Introduction Interoperability is the challenge of getting processes to share and exchange information effectively. Service orientation relates to creating self-contained, self-describing, accessible, and open, computer services. Both these challenges relate to the representation of the data being exchanged/manipulated. There are various existing sources of research information in the University, for example, ePrints – the publications repository. Research information is complex, structured data, and the future requirements of it are only partially known. If we commit to one encoding, or even one representation language, later it may turn out to be inadequate or obsolete. Current work [10] on these issues points to representing the data in an ontology. More specifically, an ontology is a notion defined by Gruber as an explicit specification of a conceptualization [8]. The term (from the Greek, ontos: of being and logia: study) is borrowed from Philosophy and it refers to the subject of existence. In Artificial Intelligence (AI), an ontology is constituted by a specific vocabulary used to describe a certain reality, plus a set of explicit assumptions regarding the intended meaning of the vocabulary [7]. Thus, the ontology describes a formal specification of a certain domain: a shared understanding of a domain of interest as well as a formal and machine understandable model of this domain. In the e-business context [6], a mechanism to improve system usability, maintenance, efficiency and interoperability could reside in the formal description of the semantic of the document-based framework for business collaborations. The formal descriptions could be provided through the definition of an ontology that represents the implicit concepts and the relationships that underlie the business vocabulary.

2. Aims and Objectives The aims of this work are: - to examine appropriate knowledge representation schemes within the context of ontologies, and an available tool support, for research information, for example, ePrints; - to ensure that the representation is consistent with the process models developed in WP2; - to develop representations of knowledge for research information in an ontological form, as identified above.

3. Methodology The creation of an ontology requires specialized skills and involves various stakeholders. The ontology development process depends on a variety of factors like the choice of the software tool used to build and edit the ontology, the language in which the ontology is

Page 24 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

implemented, the methodology which will be followed to develop it, the applications in which it will be used, the type of the ontology under construction, the available formal and informal existing knowledge resources, such as lexicons, existing ontologies, etc, and may include a large number of necessary activities. There is no an established and unique procedure to develop ontologies despite several methodologies have been proposed over time [4], [5]. However, four general tasks to build ontologies have been identified: - selection (includes selection of the available resources – related literature, existing ontologies, group of expert to the domain under description, selection of the appropriate tool and language); - analysis (includes analysis of selected resources, of the classes and the properties of the selected ontology); - definition (includes definition of what is important for the description of a specific domain through the competency questions, definition of the purpose and the domain of the ontology, the definition of the classes, the class hierarchy, the properties and the instances of the ontology); - evaluation (includes evaluation of the selected resources, evaluation of the technical quality of the ontology and evaluation of the overall quality of the obtained results). These tasks are distributed into different phases: - the specification phase (answers why the ontology is being built, what its intended uses are, who the end-users are); - the conceptualization phase (conceptualizes the domain knowledge); - the implementation phase (transforms the conceptual model into a formal computable model); - the evaluation phase (assesses the resulting ontology). These phases correspond roughly to the main steps of software engineering methodologies like the IEEE Standard 1074-1995 for Developing Software Life Cycle Processes [5]. Moreover, the ontology development methodologies could be classified into two categories: - methodologies focused on building a single ontology for a specific ontology for a specific domain of interest; - methodologies focused on the construction of ontology networks. The single ontologies could be further distinguished among those aiming at building ontologies: - from scratch; - by reusing pre-existing ontologies; - by using non-ontological resources. These single ontologies are also divided in collaborative and non-collaborative, according to the degree of participation of the involved ontology engineers, users, knowledge engineers and domain experts in the ontology engineering process. They are also described as application dependent, semi-application dependent and application independent, according to the degree of dependency of the developed ontology on its final application. The single-ontology capture approach could vary according to the adopted strategy for identifying concepts, and could be bottom-up (from the most concrete to the most abstract), top-down (from the most abstract to the most concrete), or middle-out (from the most relevant to the most abstract and most concrete).

Page 25 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

We can further distinguish manual, semi-automatic and automatic ontology construction, according to the degree of human involvement in the building process. However, the above mentioned criteria are not standards. Furthermore, none of the methodologies proposed in the literature [5] are fully mature and they need then to be adapted to the project needs. Hence, we have followed our original methodology based on the criteria previously enumerated, while developing the new ontology called ePrOnto (from the contraction of ePrints and Ontology), and we have carried out the actions as described below.

4. Implementation As noted above, we decided to investigate the application of ontology using the publications repository. Capturing this information within a standard ontology language would make it universally accessible throughout the web, and allow it to be analysed, queried and compared using powerful, open tools. We interviewed the Head of Computing Library Services of the University of Huddersfield to understand the mechanism of ePrints publication repository in order to capture ePrints knowledge to build system interoperability services. We thus identified the actors interacting with the ePrints system as well as the procedures by which the actors interact with this system. The resulting UseCase diagram (Fig. 1) uses the standard Unified Modeling Language (UML) [3] and shows the functionality of the system as well as its dependencies at a high level viewpoint.

Figure 1: UseCase Diagram describing the action of encoding an Item into ePrints

Next, we have modelled the business process using Buisness Process Modeling Language (BPML) [1] and generated the flowchart shown in Fig. 2. The UseCase diagram as well as the Buisness Process Modeling Notation (BPMN) has been designed with Modelio Free Edition v1.2 [1], mainly because this software supports UML and business modelling while being a user-friendly and free tool. These steps have helped us in answering to the competency questions to determine the domain and the scope (“what we do with it�) of the ontology.

Page 26 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 2: BPMN Diagram describing the action of adding/removing an Item in ePrints

For a Pilot implementation we developed ePrOnto, the covered domain is ePrints and its scope is to enable interoperability/sharing knowledge, efficient maintenance, and also question queries. Moreover, ePrOnto is application independent, that means that it is the same ontology e.g. for maintenance as well as for the query purpose. Then, in order to build the domain ontology, we have selected and analyzed the ePrints vocabulary. The key related words have been identified when logging into ePrints system and doing the task of adding an Item to the repository. To capture the ontology, we have chosen Protégé v4.1 [2], running on a Windows platform. This choice is motivated by the fact that this tool [9] facilities the interoperability with other knowledge-representation systems and has a user-friendly, configurable interface. The adopted language to express the ontology is the Web Ontology Language (OWL) [12], according to the World Wide Web Consortium (W3C) recommendation. In particular, we have adopted OWL-DL specie because, on one hand, it is more expressive than OWL-Lite – an OWL sub-language only adapted for simple situations. On the other hand, OWL-DL is based on Description Logics (DL). Thus, it is possible to perform automated reasoning on OWL-DL-based ontology like in [11]. That is not the case for OWL-Full-based one, where OWL-Full is union of OWL syntax and Resource Description Framework (RDF)’s data representation. Moreover, OWL-DL enables reasoner use to compute the inferred ontology class hierarchy and to perform the consistency check. To develop ePrOnto, we have identified specific basic concepts used as cornerstones for the ontology design. Next, we have mapped these concepts to a set of OWL main classes, which represent the roots of a set of corresponding subclasses together with their relationships with other classes. Each of these identified concepts represents also a starting point for the browsing of the ontology. Protégé Instances Slots Classes

OWL Individuals Properties Classes

Table I: Definition equivalence between Protégé and OWL

A Protégé ontology consists of classes, slots, facets, and axioms as mentioned in Table I. Classes are concepts in the domain of discourse and constitute a taxonomic hierarchy. Slots describe properties or attributes of classes and instances. Facets describe properties of slots. Axioms specify additional constraints. A Protégé knowledge base includes the ontology and individual instances of classes with specific values for slots.

Page 27 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

OWL ontology has similar components to Protégé-based one. However, the terminology used to describe these components is slightly different (see Table I). OWL classes are interpreted as sets that contain individuals. An example of ePrOnto class is shown in Fig. 3. The class is called Contributor and contains four subclasses (Contribution; Contributor_Email; Contributor_Family_Name; Contributor_Given_Name/_Initials).

Figure 3: Example of Class (Contributor) implemented into ePrOnto ontology

Individuals represent objects in the domain that we are interested in. Instances can be referred to as being “instances of classes”. In our example, the subclass Contribution has individuals as partially shown in Fig. 4.

Figure 4: Example of Individuals of Contribution Sub-Class implemented into ePrOnto ontology

Page 28 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Properties are binary relations on individuals – i.e. properties that link two individuals together. They can have inverses. Properties can be limited to having a single value – i.e. to being functional. They can also be either transitive or symmetric. Properties are also used to create restrictions in OWL. The latter ones could be of three categories: quantifier restrictions, cardinality restrictions, and hasValue restrictions. These quantifier restrictions are composed of a quantifier, a property, and a filler. The two quantifiers that may be used are: the existential quantifier (read as at least or some in OWL speak) and the universal quantifier (read as only in OWL speak). Hence for a set of individuals, an existential restriction ( ) specifies the existence of a (i.e. at least one) relationship along a given property to an individual that is a member of a specific class. For example, hasContribution Contributor describes all of the individuals that have at least one (some) relationship along the hasContribution property to an individual that is member of the class Contributor.

5. Outputs and Results The ontology for ePrints Information Management we called ePrOnto could be characterized in regards to the classification presented in the methodology section. Hence, ePronto ontology, developed with Protégé OWL, could be considered as a semi-automatic single ontology with multiple layers (different levels of hierarchy). Some of the layers could be seen in left part of Fig. 5 as well as the automatically generated OWL code. An overview of the whole ePrOnto ontology structure is demonstrated in the right part of Fig. 5. Moreover, ePrOnto was designed in a one-step collaborative way from scratch and is application independent. The adopted approach for the knowledge capture is a middle-out strategy. As none of the ontology development methodologies described in the literature were directly suitable, ePrOnto was developed according to a unique scheme.

Page 29 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Figure 5: Illustration of ePrOnto ontology

Some of the Items (publications) from ePrints have been encoded into ePrOnto to validate the interoperable services of our designed and implemented ontology. Hence, the proposed ontology is a first ontology developed for ePrints repository management, and which could be compatible with query process.

6. Conclusions and Perspectives This study was focused on the modelling and the efficient management of the publication repository (ePrints) system of the University of Huddersfield, in an interoperable way. Hence, the related complex data and information have been transformed into structured knowledge through the use of the ontological approach. This has lead to the design and the implementation of ePrOnto – the OWL-based ontology for ePrints information management. A very recent work [13] has proposed to automatically extract topics from text corpus. Following this direction, our future work will be the development of an innovative method to automatically update the developed ontology (ePrOnto) in order to provide a fully automatic interoperable service.

Page 30 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

7. References [1]



S. W. Ambler, The Elements of UML 2.0 Style, Cambridge University Press, New York, USA, 2005.


N. Dahlem, J. Guo, A. Hahn and M. Reinelt, “Towards an user-friendly ontology design methodology”, In Proceedings of the IEEE International Conference on Interoperability for Enterprise Software and Applications (IESA’09), pp. 180-186, April 2009.


M. Fernandez Lopez, “Overview of methodologies for building ontologies”, In Proceedings of the IJCAI Workshop on Ontologies and Problem Solving Methods (KRR5'99), pp. 4.1-4.13, August 1999.


N. Gessa, M. Busanelli, P. De Sabbata, and F. Vitali, “Extracting a semantic view from an ebuisness vocabulary”, In Proceedings of the IEEE International Conference on Enterprise Computing, E-Commerce, and E-Services (EEE’06), pp. 57-57, June 2006.


P. Green and M. Rosemann, Buisness Systems Analysis with Ontologies, Idea Group Publishing, Hershey, PA, USA, 2005.


T. R. Gruber, “Towards principles for the design of ontologies used for knowledge sharing”, International Journal of Human-Computer Studies, 43(5-6):907-928, November 1995.


M. Horridge, A Practical Guide to Building OWL Ontologies Using Protégé 4 and CO-ODE Tools, Ed. 1.2, March 2009.


S. Jain and J. Pareek, “Automatic topic(s) identification from learning material: An ontological approach”, In Proceedings of the IEEE International Conference on Computer Engineering and Applications (ICCEA'10), pp. 358-362, March 2010.


M. Peim, E. Franconi, N. W. Paton, and C. A. Goble, “Query processing with description logic ontologies over object-wrapped databases”, In Proceedings of the IEEE International Conference on Scientific and Statistical Database Management (SSDBM’02), pp. 27-36, July 2002.


S. Suwanmanee, D. Benslimane, and P. Thiran, “OWL-based approach for semantic interoperability”, In Proceedings of the IEEE International Conference on Advanced Information Networking and Applications (AINA’05), vol. 1, pp. 145-150, March 2005.


W. Wei, P. Barnaghi, and A. Bargiela, “Probabilistic topic models for learning terminological ontologies”, IEEE Transactions on Knowledge and Data Engineering, 22(7):1028-1040, July 2010.

Page 31 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

APPENDIX B - University business computer systems and their functions. System Name

What it is

Active Directory

Manages user identities, usernames and passwords

Agresso (Finance System)

Alumni Database


Communicatio ns with other systems usernames and passwords used by most other systems


Current R&E uses

Outlook address book, etc

Computerised financial system / ERP


Anything you want to know about Finance

Allows university to keep in contact with alumni AKA SITS, student management and planning

Raiser's Edge / ASIS - being migrated to ASIS PINS “Corporate Database”

R&E planning reports uses AD to link data from HR and Repository Anything related to finance (i.e. purchasing, income) This will link to the new CRM system

Timetabling Space Management Contacts Upto date list of Directory all telephone, email and room number contacts External Data re contracts Examiners’ for Taught database course External Examiners and information on all claims HR System HR data, staff grades, job roles, (Bond Professional) etc

ASIS Business Objects reports, PINS reports


At present none, but could integrate with Staff Profiles ASIS? Contracts up for renewal, ad hoc external examiner queries

Sends payroll numbers to Active Directory

Page 32 of 35

Supervision responsibilitie s and staff registered for research degrees for the R&E planning reports Possible use for capacity planning Have established link to this database None - as only used for taught courses

Staff data for the R&E planning reports

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10


Web Content Management System

Lotus Notes

There are a number of Lotus Notes applications such as “CRM”, “Committees” and “Experts” Student placements database


Can display information from Staff Profiles database, etc

Repository (ePrints)

Repository of academic outputs

Research Grants and Contracts in progress database Research & Enterprise Data Warehouse

Lists RG&C in progress, currently an Excel spreadsheet Aggregates data from RG&C database, ASIS, Repository and HR System for analysis and reporting Analytical/Reporti ng database for ASIS information

Some information is ultimately input to Agresso

Applicant Portal and Preenrolment portal websites used by Postgrads Systems integration, business intelligence, collaboration and bpm platform

ASIS, Library, LCMS, Repository

PINS Corporate Database PortalPlus


Reporting capabilities under development

Manage website content, display Staff Profiles data Lotus notes is no longer being used and is to be replaced with various other systems This system is being replaced and is no longer used by R&E Outputs counted for R&E planning reports Used for R&E planning reports

Reads from RG&C database, ASIS, Repository, HR System

R&E planning reports, Repository Reports

R&E planning reports


SQL Server Reporting Services, custom web reports Current users of the system, the number of home/international students Working to deliver R&E planning reports through SharePoint and SQL Server Reporting Services

None at present

Page 33 of 35

Research Office

Team collaboration, electronic forms, workflow and process management

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10


RIMS Overview General The RIMS [Research Information Management System] system exposes a number of different user interfaces dependant on the user classification. Users are classified as Academics, School Administrators, Repository Administrators, Super Users and the Vice-Chancellor’s Office. An individual user may be in more than one of the categories.

Academic Users Academic users when accessing the RIMS system, are presented with an individual home page. Their home page contains a lists all their research publications and where appropriate a request to supply further data about some of those publications, see Figure 1. The credentials and identity of users are checked automatically without a requirement to log in to the system. The Home page contains a statement detailing the criterion used by Research and Enterprise to count/evaluate publications. This page allows the user to check the accuracy of the information being used by central administration. One of the most common problems in accessing research publications is that not all publications are correctly assigned to the correct academic. This is often as a result of inaccurate or incomplete data that has been entered into the University repository system. Users who believe that a publication meeting the criterion specified is absent from their list may use a search facility to check if the publication has been allocated to any of the other collaborators and to check if the publication is known on the system. If a publication is not known on the system the user is encouraged to check the repository and ensure that a standard form of their name or other identifying credentials are present on the repository entry for that publication. Further information concerning ‘impact’ statements is requested for publications that fall in the period of the forthcoming Research Assessment exercise. The impact statements can be entered and edited by any contributor to the publication. The appropriate criteria to be used is specified on the input page.

School Administrators School Administrators are presented with a page to allow them to generate reports as seen by central administration on the publications allocated to staff members within their own school. This provides administrators with the information to challenge or confirm any statistics generated centrally. In addition to publication reports School Administrators can check the details of which members of staff are being attributed to the school as recorded in the HR database.

Page 34 of 35

Project Acronym: BRIM project Version: 1.0 Contact:Professor Lee McCluskey Date: 30/09/10

Repository Administrators Repository Administrators can see the statistical reports generated for schools and individual academics. They also see reports on anomalies in the data extracted from the repository. The report details authors/contributors to publications that cannot be identified with recognised members of staff nor with known students or external academics. This report tries to match up names provided in the repository with members of staff to help identify the contributors. Mismatches can be recorded to suppress such requests for identification being made again. Correct identifications should enable the administrator to update the repository with more complete information.

Vice-Chancellor’s Office The Vice-Chancellor’s Office can view all statistical reports generated on a University wide basis and to see reports generated on individual schools.

Super Users Super users have full rights over the system and can see and perform any of the actions allocated to other users, and are based within Research and Enterprise. Additionally they can control access to the system and set system parameters. Super users have control of the text describing criteria for inclusion and evaluation of publications and may change this on a by school basis.

Figure 4 Sample Academic Home Page

Page 35 of 35


Project Acronym BRIM Project Title Using Business Process Management Tools and Methods for Building Research Information Management Start Da...

Read more
Read more
Similar to
Popular now
Just for you