A MT-Oriented Approach to Entailment Recognition

Entailment is a relation between different expressions of the same meaning. If a hypothesis can be inferred from a text T, then T entails H(T→H). Entailment between two English sentences with light verb construction is studied from the viewpoint of their internal dependency structure. Entailment rule "take a break→break" is taken as an example, and one hundred dependent nouns for "take" and "break" are taken from the Corpus of Contemporary American English. Features are selected from the dependent nouns to compute similarity between "take" and "break". Ten examples are computed in total and their values fall within [1.00,1.50]. This result is used to recognize entailment between two English sentences for automatic translation or the online chat robot "Xialiao" which can return to users an answer which is entailed in the question.