Identification of discourse relations from biomedical texts and its application in cause effect extraction

Abstract

The objective of this work is to study the discourse relations in biomedical document and to develop a discourse relation identification system In this study the analysis of discourse relations established by discourse connectives is considered They are linguistic expressions signaling discourse relations and contribute to the discourse coherence They connect discourse units such as clauses or sentences and function as a primary source for identifying and describing syntactic and semantic structure of a discourse Further a deeper analysis of these relations is attempted to extract cause effect from the discourse analysed corpus A causative construction contains a causal marker cause and effect In this study the automatic extraction of cause and effect established by the causal discourse connective is considered We have also concentrated on extraction of Biomedical Named Entities BNEs like gene protein disease etc from biomedical document The task of automatically identifying and classifying these BNEs from biomedical text to predefined classes is termed as Biomedical Named Entity Recognition BNER Identification of BNEs with information of causal discourse relation would benefit the development of more sophisticated information extraction systems Our system first identifies and extracts the explicit discourse relations then identifies the BNEs and finally extracts cause and effect We have used supervised machine learning technique Conditional Random Fields and Support Vector Machine to develop the models. newline