Conversational Question Answering Over Knowledge Base using Chat-Bot Framework

摘要

Conversational Question Answering aims at answering natural language questions via well-structured relation information between entities stored in knowledge base. Knowledge Base Question Answering is one of the promising approaches for extracting substantial information from the Knowledge Bases. Existing Question Answering systems answer each Question independently, adding redundancy to repeat the entity even if the current Question is a follow-up one to the previous one. In this paper, we propose a Robust-Answer-Driven-Assistant (RADA) using the chatbot framework to overcome this problem. It consists of an ensemble of Entity Recognition, Entity Prediction, Question Answering models, and dialogue system. We conduct quantitative experiments, including comparisons with the state-of-the-art on the Web-Question dataset. Our experiments suggest the effectiveness of RADA in comparison with other methods under the F1-score metric.