Sentence extraction in recognition textual entailment task

摘要

Recognizing textual entailment (RTE) is a task that predict whether a text fragment can be inferred from another text fragment. In this paper, we tackle RTE problem using sentence extraction to cover semantic variation and then extracting subject, predicate and object from each sentence without using external resources like Wordnet. Finally, similarity function is used to predict entailment relation. In sentence extraction phase, we used sentence detection, extract sentence in subordinate clause, prepositional phrase and passive sentence. Our system has accuracy of 0.63 which is comparable to other system that is not using external resources.