Overview
Topic status: We're looking for students to study this topic.
Search engines have reached a threshold in performance and keyword spotting is still the only usable paradigm (the one used by millions when querying on Google). The field of information retrieval (IR) has developed numerous corpus-based techniques for the automatic expansion of queries. Such techniques often compute expansion terms using a selection of terms appearing in the top-ranked documents in relation to the initial query.
This project aims at using human associations to expand queries, on single words and on compounds. The first step will be to compare human association versus state of the art information retrieval systems (such as Lemur) via various models for query. The second step will consist in using automatic approaches inspired from human cognition to improve query expansion in the same way as human data.
Hypothesis/Aims
This project assumes that cognitive processes underpin choices about which terms users' select in order to expand their queries. For this reason, human word association data will be employed as a means for producing useful expansion terms particularly for those queries which have thus far resisted performance improvement. Suits: IT student with a background in programming.
Approaches
- Use and improvement of existing search engines
- Design and implementation of a query expansion component
- Empirical evaluation using standard IR test collections and performance measures
References
- Song, D. and Bruza, P.D. (2003) Towards context-sensitive information inference. Journal of the American Society for Information Science and Technology, vol. 54, no. 3, pp. 321-334
- Study level
- Honours
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Contact
- Please contact the supervisor.
Professor Peter Bruza
Dr Laurianne Sitbon
