Pseudo-Relevance Feedback for Information Retrieval in Medicine Using Genetic Algorithms

摘要

Pseudo-Relevance Feedback is one of the methods for improving search engine results. By automatically extracting information from a previous search result, a new query is posed as an expansion of the original query, and then it is searched again. In this paper, we apply a genetic algorithm to improve the Pseudo-Relevance Feedback method in searching medical texts. First, a set of candidate terms is constructed by extracting keywords from the documents returned from the initial search using the original query. Then, the seed terms are selected from the candidate term set using our proposed genetic algorithm, to be merged with the original query to create a new query. The new query is searched again, returning a final ranked list of documents. Experimental results on the TREC 2014 CDS dataset show that the proposed method outperforms the baseline method that does not use a genetic algorithm for Pseudo-Relevance Feedback.