A Multi-Concept Semantic Representation System for Chinese Intent Recognition

Addressing the complex language understanding challenges in task-oriented dialogues, we propose a multiconcept semantic representation system specifically designed for Chinese intent recognition, aimed at providing a unified solution to the difficulties encountered in intent recognition. We describe the Chinese multi-concept semantic representation system and detail the initial construction of a Chinese multi-concept semantic annotation corpus, using both the Housing Fund intent dataset and the query classification dataset. This corpus was subject to a statistical analysis of its annotations. The ultimate goal of this research is to establish a cross-domain, transferable multi-concept semantic representation system tailored for Chinese intent recognition. Additionally, this paper aims to develop a corpus and knowledge base for training multi-concept semantic annotation models across multiple domain scenarios, thereby providing a theoretical and data-driven foundation for language understanding focused on Chinese intent recognition.