The Role and Resolution of Textual Entailment in Natural Language Processing Applications

摘要

A fundamental phenomenon in Natural Language Processing concerns the semantic variability of expressions. Identifying that two texts express the same meaning with different words is a challenging problem. We discuss the role of entailment for various Natural Language Processing applications and develop a machine learning system for their resolution. In our system, text similarity is based on the number of consecutive and non-consecutive word overlaps between two texts. The system is language and resource independent, as it does not use external knowledge resources such as WordNet, thesaurus, semantic, syntactic or part-of-speech tagging tools. In this paper all tests were done for English, but our system can be used with no restrains by other languages.