Memory Efficient Knowledge Base Question Answering with Chatbot Framework

The goal of Knowledge-Based Question Answering with Chatbot Framework is to target and improve the ability of chatbot’s existing framework to better recognize and answer informal and natural language patterns via well-structured relational information between entities stored in knowledge bases. Knowledge extraction is not always a one-step process as there may be follow-up questions to the previous answer. The flaw in existing Question Answering systems is that they answer each question independently, adding redundancy for the user to repeat their previous question even if a follow-up was already asked. In this research paper, we are proposing a Knowledge Base Question Answering (KB-QA) system using a chatbot framework to tame this problem. Our approach of KB-QA is capable of conversing with live users utilizing the context of the user’s interaction with the Knowledge Base in the current session. Our approach consists of an ensemble of entity resolution, entity prediction, question answering, and co-reference resolution models. We have conducted quantitative experiments and comparisons with existing state-of-the-art methods, on the Web-Question dataset. Using an F1-score metric, the effectiveness of our memory-efficient approach is evident from the results when evaluated with other conventional methods