Optimizing a model for library intelligent question-answering system through constructivist theory lens

Purpose This study aims to create an optimized ChatGLM model – JG-ChatGLM for the library intelligent question-answering (IQA) system through a constructivist theory lens to improve the professional ability and communicative competence of it. Design/methodology/approach This study uses ChatGLM as an algorithm base, takes LangChain as a framework and integrates constructivism theories to construct JG-ChatGLM as the process of corpus feeding, constructing, fine-tuning and evaluating. Findings The oversensitivity and overstability of the new model JG-ChatGLM has significantly improved compared with ChatGLM and ChatGPT-3.5. Research limitations/implications JG-ChatGLM’s stability and maturity still need to be further enhanced; also, professional corpus gathering and annotation across various disciplines still face some difficulties. Originality/value JG-ChatGLM may create an online community in the IQA system, recognize multimodal interaction and offer individualized and varied information services for various groups of individuals, which addresses the shortcomings of the library IQA system at this stage and meets the expectations outlined in prior studies. It also provides a reference for library IQA system construction, expands the constructivism theory application and deepens the theory about library service.