Application of large language model in intelligent Q&A of digital government

LLM (Large Language Model) is developing rapidly today, and it is very important to use LLM to help existing question answering systems improve. The government question answering system is of great value in improving government administrative efficiency. Existing intelligent questions and answers mostly use the combination of keyword matching and machine learning. But keyword matching is difficult to understand complex and multi-round dialogue questions, resulting in limited quality of reply content. Based on machine learning method, it has further improved the understanding of semantics, but because the understanding of words in the field of government affairs is too difficult and professional; and the semantic recognition of colloquial questions is difficult, the performance is not even as good as the question answering system based on keyword matching. This paper uses the large language model as a tool to help understand user questions, and integrates it into the existing government question answering system. On the premise of effectively utilizing the advantages of the large language model, it avoids its defects. And it is demonstrated by experiments that the above system has a great improvement compared with the previous method.