Overview
Topic status: We're looking for students to study this topic.
Search engines exhibit different performance quality with respect to different user information needs. For instance, the most popular web search engines perform well in document retrieval mode, but are not very good at identifying the specific location of content within large documents. Some search engines specialize in Question Answering, others specialize in domain specific knowledge. There is a need to evaluate search engine performance in an objective manner with respect to desirable behaviour and user needs. In this project we are interested in the evaluation of search engines in Passage Retrieval. The task is to return passages within documents that are relevant as answers to queries. Rather than develop a search engine, the task is to take a set of queries, a set of results-sets from several search engines, a set of relevant passages that satisfy the queries. The evaluation program has to derive a performance score for each result-set based on how well it is able to match the known relevant passages. We are dealing with the Wikipedia collection of articles, in XML format.
The evaluation strategy has to be explored and a program should be implemented in Java that performs the evaluation in an effective manner. XML processing will be required. Good programming skills and an aptitude for writing efficient programs are required.
- Study level
- Honours
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Contact
-
Please contact the supervisor.
Associate Professor Shlomo Geva
