Using Role Determination and Expert Mining in the Enterprise Environment
Abstract
There are two traditional ways of doing an expert search: either we look for the desired topic in every candidate's documents, or we look into every document that contains information on the topic and try to find any experts whose name appears there. In the real world, expert search is not only name matching. Since each expert has his/her own characteristics, we try two methods to judge whether the person we have found is likely to be an expert. One method is to determine the role of a person by the context of the pages; the other is to judge the authority of a person by the forms of pages where he/she appears (e.g., resumes on an intranet). For the expert search task this year, we evaluate candidates from three aspects: relevant page retrieval, role determination, and expert page mining. The results show that these methodologies have been helpful in improving the performance of the expert search on TREC 08 queries.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2008
- Accession Number
- ADA512740
Entities
People
- Jing Yao
- Jun Xu
- Junyu Niu
Organizations
- Fudan University