Recognizing Textual Entailment with Temporal Expressions in Natural Language Texts

Abstract

The TACTE system proposed in this paper focuses on one problem in natural language processing, namely recognizing textual entailment involving temporal expressions. The system consists of two components: one for temporal expression extraction and anchoring, and the other one for recognizing textual entailment based on events. The entailment rules are constructed using a small set of temporal expression relations and lexical resources. Several experiments are conducted, and various aspects of the system performance are illustrated. The evaluation on different data sets shows the great improvement of our TACTE system in comparison with the baseline. As a system potentially to be integrated into a larger framework, TACTE is shown to be very promising as a specialized module on entailment cases where temporal expression information is available.