NIL_UCM: Extracting Drug-Drug interactions from text through combination of sequence and tree kernels

Second Joint Conference on Lexical and Computational Semantics (*SEM), Volume 2: Seventh International Workshop on Semantic Evaluation (SemEval 2013), pages 644–650, Atlanta, Georgia, June 14-15, 2013. c⃝2013 Association for Computational Linguistics NIL UCM: Extracting Drug-Drug interactions from text through combination of sequence and tree kernels Behrouz Bokharaeian, Alberto D´ıaz Natural Interaction Based on Language Group Universidad Complutense de Madrid Madrid 28011, Spain {bokharaeian, albertodiaz}@fdi.ucm.es Abstract A drug-drug interaction (DDI) occurs when one drug affects the level or activity of another drug. Semeval 2013 DDI Extraction challenge is going to be held with the aim of identify- ing the state of the art relation extraction algo- rithms. In this paper we firstly review some of the existing approaches in relation extraction generally and biomedical relations especially. And secondly we will explain our SVM based approaches that use lexical, morphosyntactic and parse tree features. Our combination of sequence and tree kernels have shown promis- ing performance with a best result of 0.54 F1 macroaverage on the test dataset. 1 Introduction A drug-drug interaction occurs when one drug af- fects the level or activity of another drug, for in- stance, drug concentrations. This interaction can result on reducing its effectiveness or possibly in- creasing its side effects (Stockley, 2007). There are some helpful DDIs but most of them are danger- ous (Aronson, 2007), for example, patients that take clarithromycin and glibenclamide concurrently may experiment hypoglycaemia. There is a great amount of information about DDI described in papers that health experts have to con- sult in order to be updated. The development of tools for extracting this type of information from biomed- ical texts would produce a clear benefit for these pro- fessionals reducing the time necessary to review the literature. Semeval 2013 DDI Extraction challenge decided to being held with the aim of identifying the state of the art algorithms for automatically extract- ing DDI from biomedical articles. This challenge has two tasks: recognition and classification of drug names and extraction of drug-drug interactions. For the second task, the input corpus contains annota- tions with the drug names. A previous Workshop on Drug-Drug Interaction Extraction (Segura-Bedmar et al., 2011) was held in 2011 in Huelva, Spain. The main difference is that the new challenge includes the classification of the drug-drug interactions in four types depending on the information that is described in the sentence making the task much more complicated than be- fore. Additionally the current task involves DDIs from two different corpora with different character- istics (Segura-Bedmar et al., 2013). We participated in the task of extracting drug-drug interactions with two approaches that exploit a rich set of tree and sequence features. Our implemented methods utilize a SVM classifier with a linear ker- nel and a rich set of lexical, morphosyntactic and se- mantic features (e.g. trigger words) extracted from texts. In addition some tree features such as shortest path and subtree features are used. 2 Related work Due to the importance of detecting biological and medical relations several methods have been …

Content truncated to 500 words