Fine-tuning Large Language Models to Provide Accurate Recognition of User Intent for Aviation Q&A Systems

User intent recognition is a critical component in intelligent Question & Answering (Q&A) systems, enabling devices like Xiaomi’s Xiao’ai assistant, Baidu’s Xiaodu assistant, and Alibaba’s Tmall Genie to precisely interpret user commands. These systems adeptly discern user inputs to facilitate various inquiries, ranging from flight and medical information to legal and shopping services, as well as municipal office-related queries. An effective Q&A system must accurately identify the user’s underlying intent to deliver accurate responses, thereby meeting customer needs and potentially reducing operational labor costs for companies. The challenge of accurately comprehending the nuances in user-generated text is significant; the precision of intent recognition directly influences the intelligence of the Q&A system. This research introduces an intent recognition model, meticulously crafted within the PaddleNLP framework, with the objective of refining the accuracy of intent detection for airline ticket booking systems. And provide stable technical support for the built smart Q&A system.