Seminář: Word2vec-based system for recognizing faceted textual entailment

The task of recognizing textual entailment (RTE) consists, roughly said, in deciding whether a given sentence, called hypothesis, can be inferred from a given coherent text. RTE belongs to the most challenging topics in NLP, with potentially myriads of applications including multidocument summarization, plagiarism detection or semantic textual similarity detection. In this talk we propose a system (joint work with Vincent Kríž) for recognizing faceted textual entailment – which can be viewed as a modification of the original RTE task – based on the popular word2vec model. We will discuss the preliminary results and ideas for further improvement.