Conversational machine reading comprehension based on cosine

With the development of natural language processing technology and artificial intelligence technology, more and more industrial applications have landed. Our lives have also been improved through intelligent voice assistants. In life, we can control various home appliances by talking to the voice assistant, and we can also chat with the AI voice assistant. In the final analysis, these activities are actually the expansion of the dialogue system, and in the dialogue system, historical dialogue information plays a vital role. The contribution of this article is to use the cosine similarity to compare the current question with the chat information that occurred before. If the similarity is high, it is added to the sentence embedding of the final model input. The advantage of this is that it improves the accuracy of the machine in answering questions in multiple rounds of conversations.