Dialogue System

With the rapid development of Internet technology and the increasing popularity of intelligent robots and intelligent hardware devices, traditional keyword-based search engine information retrieval is no longer sufficient. Dialogue systems that enable users to interact naturally with robots and devices through voice dialogue and other natural language methods have gained attention and have a profound impact on daily life. Dialogue systems can be non-task-oriented or task-oriented, and can save labor costs and improve work efficiency by replacing repetitive mental work. However, building a human-computer dialogue system is challenging, especially the natural language understanding module that converts user input into structured semantic representations. Discovering unknown user intents that have not appeared in the training set and have not been identified is crucial for improving service quality and providing personalized services. Existing research on new intent discovery is divided into three schools based on unknown intent detection, non-clustering, and semi-clustering, but accurately identifying and understanding new types of user intents is still challenging. Efficiently identifying intents is a key step in answering user questions, and continuous optimization of intent recognition and discovery algorithms is necessary. The discovery of new intents in dialogue is a promising research field with high commercial value.