A Fine-grained Interactive Question Answering Model Fused with Knowledge Graph Information
The existing answer selection model fused with knowledge graph compresses the words in the question and answer pairs and the corresponding knowledge entities into two vectors through the attention mechanism,and then calculates the matching degree of the question and answer pairs,which does not make good use of the fine-grained interaction information between ques-tion-answer pairs.Therefore,this paper proposes a fine-grained interactive question answering model that integrates knowledge graph information.From the perspectives of sentence-level coarse-grained and word-level fine-grained,which makes better use of fine-grained information,thereby improving the matching effect of the model.Finally,the effectiveness of the model is verified on the general English dataset TrecQA and the Chinese medical dataset cMedQA2.
deep learningnatural language processingquestion and answer matchingknowledge graph
万方数据
维普
