Let your users do the ranking.
Ranking search results is a thorny issue for enterprise search. Search engines rank results using a variety of sophisticated algorithms, but users still complain that search can't ever seem to find anything useful or relevant! The challenge is to provide results that are ranked according to the users' definition of relevancy. Sandia National Laboratories has enhanced its commercial search engine to discover user preferences, re-ranking results accordingly. Immediate positive impact was achieved by modeling historical data consisting of user queries and subsequent result clicks. New data is incorporated into the model daily. An important benefit is that results improve naturally and automatically over time as a function of user actions. This session presents the method employed, how it was integrated with the search engine,metrics illustrating the subsequent improvement to the users' search experience, and plans for implementation with Sandia's FAST for SharePoint 2010 search engine.
- Research Organization:
- Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 1035664
- Report Number(s):
- SAND2010-8542C; TRN: US201205%%229
- Resource Relation:
- Conference: Proposed for presentation at the Enterprise Search Summit held May 10-11, 2011 in New York, NY.
- Country of Publication:
- United States
- Language:
- English
Similar Records
DOE SBIR Phase II Final Report: Distributed Relevance Ranking in Heterogeneous Document Collections
Streamlining Stakeholder Data Accessibility Utilizing Web-Based GIS Capabilities - 18672