Mkgqa: Multi-Turn Question Answering System Based on Medical Knowledge Graph

The development of question answering for knowledge graph (KGQA) and the emergence of the graph database such as NEO4J allow to increase performance on domain-specific question answering system (QAs). However, it is hard to understand questions especially in the step of semantic analysis. This research is focused on the multi-turn QAs in the medical field. Our contribution in this paper is twofold: (1) we introduce a pre-trained language model for classification questions intentions. This model attempt to divides questions raised by users into several categories, and (2) we present a neural model for named entity recognition (NER). The aim of this model is to identify the common vocabulary in the medical field. An additional advantage of above two models can support multi-turn question answering (QA) by recycling the model results. From this study, we conclude that the proposed models work well and give accurate estimates in KGQA.