Chinese Textual Contradiction Recognition Based On Linguistic Phenomena

Text entailment is foundational for text understanding and semantic inference inorder to solve the problem of variability of semantic expressions in natural language,which has recently received a surge of interest in the computational linguisticscommunity. Text contradiction is the negation of text entailment. The standardarddescription for the term ‘textual contradiction’is to consider text T and H contradictory ifthere is no possible world in which T and H are both true. Detecting contradictive texts isa crucial and fundamental work for text understanding just like textual entailment andboth of them are the key aspects of semantics.This paper focuses on textual contradiction detection on the basis of textualentailment. A deep analysis on contradiction phenomena including quantity exclusion,temporal exclusion, spatial exclusion, modifier exclusion, antonym and negation isconducted to identify contradictory text pairs.In this paper, the Chinese textual contradiction approach based on linguisticphenomena has been put forward. After using linguistic phenomena analysis,contradiction linguistic phenomena are converted to semantic rules and sematic features.Then contradiction optimization methods based on semantic rules and two-stageclassificaton are adopted to optimize the classification results by SVM classifier.In the evaluation of the experimental results, we use three common standards,Precision, Recall and F1-measure. The experiment results demonstrate the effectivenessand feasibility of the textual contradiction recognition approach based on linguisticphenomena which can also significantly improve the performance of textual entailmentrecognition.