ABSTRACT
Text Entailment relation between two texts: T (the text) and H (the hypothesis) represents a fundamental phenomenon of natural language. It is denoted by T ‚ H and means that the meaning of H can be inferred from the meaning of T (or T entails H). Even this defi nition suggests that the Text Entailment (TE) problem concerns a semantic aspect (the word meaning) and a logical aspect (the word inference). Indeed, the recognition of textual entailment is one of the most complex tasks in natural language processing and the progress on this task is the key to many applications such as Question Answering, Information Extraction, Information Retrieval, Text summarization, and others. For example, a Question Answering (QA) system has to identify texts that entail the expected answer. Given a question (Q), and turning this into an correct answer (Q'), the text found as answer (A) entails the expected answer (Q'), (A ‚ Q'). Similarly, in Information Retrieval (IR) the concept denoted by a query expression (Q) should be entailed by relevant documents retrieved (R), (R ‚ Q). In text automated summarization or (TS) a redundant sentence or expression (R), to be omitted from the summary (S), should be entailed by other sentences in the summary (S ‚ R). In Information Extraction (IE) entailment holds between different text variants that express the same target extraction. In Machine Translation (MT) a correct translation should be semantically equivalent to the standard translation, and thus both translations have to entail each other. Let it be remarked that the paraphrasing problem is in fact related with TE, as a bidirectional entailment: T1 is a paraphrase of T2 if T1 entails T2 and T2 entails T1. Thus, in a similar way with Word Sense Disambiguation which
is recognized as a generic task, solving textual entailment may consolidate the performance of a number of different tasks of NLP (QA, IR, MT, IE, TS).
