关键词: 问答模型, 文本分类, 文本匹配

Abstract: Text classification and text matching are two ways to solve the frequently asked questions (FAQ)-based question answering. However, using the text classification method alone cannot effectively utilize the information of the standard question itself. When the text matching method is used alone, the selection of the negative sample is a very difficult problem. A series of models that combine text classification and text matching methods are proposed, which not only can select negative examples that really need to be distinguished, but also can effectively use the information of standard questions. Experiments show that the proposed models achieve the best performance on multiple FAQ-based question answering data.

Key words: question answering, text classification, text matching