Mining linguistic cues for query expansion: applications to drug interaction search

  • Authors:
  • Sheng Guo;Naren Ramakrishnan

  • Affiliations:
  • Virginia Tech, Blacksburg, VA, USA;Virginia Tech, Blacksburg, VA, USA

  • Venue:
  • Proceedings of the 18th ACM conference on Information and knowledge management
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Given a drug under development, what are other drugs or biochemical compounds that it might interact with? Early answers to this question, by mining the literature, are valuable for pharmaceutical companies, both monetarily and in avoiding public relations nightmares. Inferring drug-drug interactions is also important in designing combination therapies for complex diseases including cancers. We study this problem as one of mining linguistic cues for query expansion. By using (only) positive instances of drug interactions, we show how we can extract linguistic cues which can then be used to expand and reformulate queries to improve the effectiveness of drug interaction search. Our approach integrates many learning paradigms: partially supervised classification, association measures for collocation mining, and feature selection in supervised learning. We demonstrate compelling results on using positive examples from the DrugBank database to seed MEDLINE searches for drug interactions. In particular, we show that purely data-driven linguistic cues can be effectively mined and applied to realize a successful domain-specific query expansion framework.