Entity Resolution of Institutions in Bibliographic Databases

Acknowledgements Many people have assisted me in carrying out this project. Firstly I would like to thank my academic supervisors, Associate Professor Peter Christen and Dr. Qing Wang for their ideas, support, encouragement and feedback. I would also like to thank Dr. Paul Wong from the ANU Research Office for providing me with a place to work and helpful advice on the project itself and the SCOPUS database. I would like to thank my friends, in particular Swapnil Mishra and Anish Varghese for their good humour and for helping me to keep things in perspective. Lastly, I would like to thank my family for their support and encouragement, and for being so understanding about all the family dinners I missed.Abstract Bibliographic databases are very important for a variety of tasks including measuring research output of institutions and for predicting future areas of research interest. However, incorrect or incomplete data in such databases can compromise any analysis and lead to poor decision making and financial loss. In this project we have performed data matching of institution data in the SCOPUS Bibliographic Database. We used a variety of established data matching methods and adapted them to the suit the particulars of the project. We describe our data cleaning work, including our novel approach for extracting institution names from the values of the organization attribute. We describe the data matching that we have undertaken, both in merging institutions where they have different identifiers but represent the same institution, and in assigning an identifier to records without one. We show that in the first case we can achieve a high coverage and maintain precision over 85%. In the second case the precision drops significantly beyond 40% coverage and we examine reasons why this is occurs. Finally, we present our conclusions along with some suggestions on how this work could be extended in the future.-4