Continuous Factual Knowledge Learning in Dialogues
摘要
Lifelong or continual knowledge learning in an interactive environment is a key capability of human beings. One can only learn so much by being told or supervised because the world is simply too complex to be completely learned this way. In fact, we humans probably learn a great deal of our knowledge through interactions with other humans and the environment around us which constantly give us explicit and implicit feedback. In the context of chatbots, continual and interactive knowledge learning is critical because in order for a chatbot to be truly intelligent in human-machine conversation, it has to continually learn new knowledge and improve itself over time. This chapter discusses works that aim to build interactive continual knowledge learning systems for chatbots with the focus on learning and extracting factual and lexical knowledge.