SemEval-2013 Task 9 : Extraction of Drug-Drug Interactions from Biomedical Texts (DDIExtraction 2013)

Second Joint Conference on Lexical and Computational Semantics (*SEM), Volume 2: Seventh International Workshop on Semantic Evaluation (SemEval 2013), pages 341–350, Atlanta, Georgia, June 14-15, 2013. c⃝2013 Association for Computational Linguistics SemEval-2013 Task 9 : Extraction of Drug-Drug Interactions from Biomedical Texts (DDIExtraction 2013) Isabel Segura-Bedmar, Paloma Mart´ınez, Mar´ıa Herrero-Zazo Universidad Carlos III de Madrid Av. Universidad, 30, Legan´es 28911, Spain {isegura,pmf}@inf.uc3m.es, mhzazo@pa.uc3m.es Abstract The DDIExtraction 2013 task concerns the recognition of drugs and extraction of drug- drug interactions that appear in biomedical literature. We propose two subtasks for the DDIExtraction 2013 Shared Task challenge: 1) the recognition and classification of drug names and 2) the extraction and classification of their interactions. Both subtasks have been very successful in participation and results. There were 14 teams who submitted a total of 38 runs. The best result reported for the first subtask was F1 of 71.5% and 65.1% for the second one. 1 Introduction The definition of drug-drug interaction (DDI) is broadly described as a change in the effects of one drug by the presence of another drug (Baxter and Stockely, 2010). The detection of DDIs is an im- portant research area in patient safety since these in- teractions can become very dangerous and increase health care costs. Drug interactions are frequently reported in journals, making medical literature the most effective source for their detection (Aronson, 2007). Therefore, Information Extraction (IE) can be of great benefit in the pharmaceutical industry al- lowing identification and extraction of relevant in- formation on DDIs and providing an interesting way of reducing the time spent by health care profession- als on reviewing the literature. The DDIExtraction 2013 follows up on a first event organized in 2011, DDIExtraction 2011 (Segura-Bedmar et al., 2011b) whose main goal was the detection of drug-drug interactions from biomedical texts. The new edition includes in addition to DDI extraction also a supporting task, the recognition and classification of pharmacologi- cal substances. DDIExtraction 2013 is designed to address the extraction of DDIs as a whole, but di- vided into two subtasks to allow separate evaluation of the performance for different aspects of the prob- lem. The shared task includes two challenges: • Task 9.1: Recognition and classification of pharmacological substances. • Task 9.2: Extraction of drug-drug interactions. Additionally, while the datasets used for the DDIExtraction 2011 task were composed by texts describing DDIs from the DrugBank database(Wishart et al., 2006), the new datasets for DDIExtraction 2013 also include MedLine abstracts in order to deal with different types of texts and language styles. This shared task has been conceived with a dual objective: advancing the state-of-the-art of text- mining techniques applied to the pharmacological domain, and providing a common framework for evaluation of the participating systems and other re- searchers interested in the task. In the next section we describe the DDI corpus used in this task. Sections 3 and 4 focus on the de- scription of the task 9.1 and 9.2 respectively. Finally, Section 5 draws the conclusions and future work. 2 …

Content truncated to 500 words