Question Answering over Knowledge Graph (KG-QA) becomes a convenient way to interact with the prevailing information. The user's information needs, i.e., input questions become more complex. We find that the comparison, relation, and opinion questions are witnessed a significant growth, especially in some domains. However, most of the current KG-QA methods cannot appropriately handle the inherent complex relation and coverage characteristics within the questions. In this work, we propose to utilize the relation information with the questions and knowledge graph in a mutual way, improving the final question answering performance. Wse design local and global attention models for relation detection. We combine the features for relation detection in an attention matching model. Experiments on our new dataset and common dataset reveal its advantages both in accuracy and efficiency.
展开▼
