Textual Entailment Search Task: An Initial Approach Based on Coreference Resolution

In this work we present our initial approach to the Recognizing Textual Entailment Search Pilot Task proposed by NIST. We proposed a new algorithm to address Text Entailment task to a document level making use of coreference resolution and then reducing this problem to a traditional main task problem. We also applied machine learning algorithms and a combination of datasets for the textual entailment task. The features chosen quantity lexical, syntactic and semantic level matching between text and hypothesis sentences. Despite our system being very preliminary, it is placed third among other existing systems.