RESEARCH ON THE CONSTRUCTION OF A KNOWLEDGE BASE FOR NEW CHINESE WORDS BASED ON LARGE LANGUAGE MODELS

The continuous emergence of new Chinese words and expressions in digital communication presents important challenges for computational lexicography, natural language processing, and knowledge organization. Conventional lexicographic resources cannot always reflect newly emerging lexical forms, evolving meanings, and changes in usage across digital communities. Although large language models (LLMs) offer substantial capabilities for contextual interpretation and structured information extraction, their application to lexical knowledge-base construction requires mechanisms for evidence verification, temporal representation, and linguistic quality control. This study proposes a provenance-aware framework for constructing a dynamic knowledge base of Chinese neologisms by integrating corpus-based candidate discovery, LLM-assisted lexical analysis, retrieval-grounded verification, and temporal knowledge representation. The framework distinguishes new lexical forms from new meanings of existing words, proper names, and non-lexical expressions. A structured knowledge model is developed to represent lexical forms, senses, formation mechanisms, contextual examples, source evidence, and temporal attributes. Additionally, a reproducible evaluation protocol is specified to assess lexical identification, semantic extraction, factual grounding, knowledge-based consistency, and updating performance. The methodological analysis demonstrates how lexical knowledge can be organized into verifiable, temporally qualified records rather than isolated machine-generated definitions. The proposed approach provides a systematic foundation for developing continuously updated Chinese lexical resources and identifies the empirical validation required before deployment.