An example-based question answering (QA) is a robust and practical approach for a real-environment information guidance system. However, it cannot appropriately respond to unexpected user’s utterances if a similar example of a questionanswer pair does not exist in the QA database; in addition, the answer sentences cannot reflect differences in nuance, because the set of answer sentences are fixed beforehand. To deal with these problems, we propose a new method, which introduces statistical machine translation training to answer sentence generation. In the proposed method, we treat questions and answer sentences as different languages. In this paper, we investigate a feasibility of translation from question into answer using real user utterances for Takemaru-kun.
展开▼