Using Odds Ratios to Extract Drug-Drug Interaction (DDI) from Biomedical Literature

• Main Authors: Ching-Chun Tu, 塗敬群
• Other Authors: Tzu-Ya Weng
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/bd5e6e

碩士 === 元智大學 === 生物與醫學資訊碩士學位學程 === 103 === Drug-Drug Interaction (DDI) indicates that two arbitrary drugs might have some interactions between each othe. We already have great amount of biometrics research on DDI analysis, so we are looking forward to providing a method besides the biometric ones. Association for Computational Linguistics (ACL) once suggested a task named "Extraction of Drug-Drug Interactions from BioMedical Texts" in 2013. The goal is to use method of Information Retrieval (IE), extract DDI interaction from biomedical data and classify them. Team FBK-irst got a F-score of 80.0% on detection, and got a F-score of 65.0% on classification. We use the same dataset of DDI Extraction 2013 and modify the Odds Ratio (OR) of evaluating variables' effectiveness in clinical trial to corpus-based features. The experiment has an F-score of 76.9% in detection, and 65.3% in classification. Though the performance is still worse than team FBK-irst in DDI detection, we have similar performance with FBK-irst in DDI classification. We provide a method to detect DDI in IE method, and it's proved to be useful. In the future, we hope to make an application based on this method in other area of IE research.