Hedge Classification in Biomedical Texts with a Weakly Supervised Selection of Keywords

Proceedings of ACL-08: HLT, pages 281–289, Columbus, Ohio, USA, June 2008. c⃝2008 Association for Computational Linguistics Hedge classification in biomedical texts with a weakly supervised selection of keywords Gy¨orgy Szarvas Research Group on Artificial Intelligence Hungarian Academy of Sciences / University of Szeged HU-6720 Szeged, Hungary szarvas@inf.u-szeged.hu Abstract Since facts or statements in a hedge or negated context typically appear as false positives, the proper handling of these language phenomena is of great importance in biomedical text min- ing. In this paper we demonstrate the impor- tance of hedge classification experimentally in two real life scenarios, namely the ICD- 9-CM coding of radiology reports and gene name Entity Extraction from scientific texts. We analysed the major differences of specu- lative language in these tasks and developed a maxent-based solution for both the free text and scientific text processing tasks. Based on our results, we draw conclusions on the pos- sible ways of tackling speculative language in biomedical texts. 1 Introduction The highly accurate identification of several regu- larly occurring language phenomena like the specu- lative use of language, negation and past tense (tem- poral resolution) is a prerequisite for the efficient processing of biomedical texts. In various natural language processing tasks, relevant statements ap- pearing in a speculative context are treated as false positives. Hedge detection seeks to perform a kind of semantic filtering of texts, that is it tries to sep- arate factual statements from speculative/uncertain ones. 1.1 Hedging in biomedical NLP To demonstrate the detrimental effects of specula- tive language on biomedical NLP tasks, we will con- sider two inherently different sample tasks, namely the ICD-9-CM coding of radiology records and gene information extraction from biomedical scientific texts. The general features of texts used in these tasks differ significantly from each other, but both tasks require the exclusion of uncertain (or specula- tive) items from processing. 1.1.1 Gene Name and interaction extraction from scientific texts The test set of the hedge classification dataset 1 (Medlock and Briscoe, 2007) has also been anno- tated for gene names2. Examples of speculative assertions: Thus, the D-mib wing phenotype may result from de- fective N inductive signaling at the D-V boundary. A similar role of Croquemort has not yet been tested, but seems likely since the crq mutant used in this study (crqKG01679) is lethal in pupae. After an automatic parallelisation of the 2 annota- tions (sentence matching) we found that a significant part of the gene names mentioned (638 occurences out of a total of 1968) appears in a speculative sen- tence. This means that approximately 1 in every 3 genes should be excluded from the interaction detec- tion process. These results suggest that a major por- tion of system false positives could be due to hedg- ing if hedge detection had been neglected by a gene interaction extraction system. 1.1.2 ICD-9-CM coding of radiology records Automating the assignment of ICD-9-CM codes for radiology records was the subject of a shared task 1http://www.cl.cam.ac.uk/∼bwm23/ 2http://www.cl.cam.ac.uk/∼nk304/ 281 challenge organised in Spring 2007. The …

Content truncated to 500 words