ABSTRACT
In the Recognizing Textual Entailment (RTE) challenge, a system has to determine whether a given Text 1 (H) entails a given Text 2 (T). Based on the task guidelines (Seventh TEXTUAL ENTAILMENT CHALLENGE@TAC 2011 MAIN TASK and NOVELTY DETECTION SUBTASK TASK Guideline), T entails H if H can be inferred from T. For example, if “ABC News emphasized that JENNINGS would remain its evening news anchor” is given as T1 and “JENNINGS was anABC news anchor” is given asT2, the system should give the answer “T1 entails T2.” In contrast, if “ABC News says JENNINGS is diagnosed with lung cancer” is given as T1 and “JENNINGS was the anchorman of ABC’s evening newscast” as T2, the system should give the answer “T1 doesn’t entail T2.” The textual entailment challenge has attracted much attention in recent years due to its potential applications such as Document Summarization, Question Answering, and Information Retrieval. In Question Answering, answers of questions can be detected based on semantic relatedness while absorbing surface difference of texts. In Document Summarization, we can retain necessary information by filtering redundant texts using RTE technologies (Watanabe et al., 2013).
