Three Questions about Clinical Information Retrieval
Abstract
Electronic Medical Records (EMRs) have greatly expanded the potential for the evidence-based improvement of clinical practice by providing a data source for computable medical information. The Text REtrieval Conference 2012 Medical Records Track (TREC-med) explored how information retrieval may support clinical research by providing an efficient means to identify cohorts for clinical studies. A shared task called participants to find cohorts of relevant patients for 50 different topic queries. The users in TREC-med information retrieval systems would be medical experts who are searching for cohorts. In our previous work, we have collaborated with such experts on specific queries; the assortment of 50 queries makes this competition a standardized benchmark task. Thus, techniques that have shown case-by-case improvement can be tested against a much larger number of queries. We have taken this opportunity to investigate three core questions around which many of our algorithms are designed: 1. What is the relative value of structured data (e.g., fields in EMRs, or document metadata) compared to clinical text? 2. Are extensive information extraction (IE) efforts any benefit when we consider the applied question of information retrieval (IR)? 3. Can distributional semantics help supply missing information in a query?
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2012
- Accession Number
- ADA581263
Entities
People
- Hongfang Liu
- James Masanz
- K. E. Ravikumar
- Stephen Wu
Organizations
- Mayo Clinic