Abstract: 【Purpose/significance】Scholar's research interest is the key feature of scholar's profile. By identifying the changing pro⁃ cess of scholar's research interest, this study can help complete academic resumes, which is of great significance for constructing com⁃ plete scholars' portraits and accurate talent discovery for cutting-edge needs.【Method/process】A text corpus comprising papers from computer science was constructed to train the BERTopic topic model for mining research topics and identifying the characteristics of scholars' research interests. Additionally, a Prompt was created to extract subject words using LLM, which were then combined with the analysis results from the topic model to describe scholars' research interests.【Result/conclusion】Compared to the baseline model, our fusion model achieved an average relative increase of 8.2% in ROUGE score and 4.5% in BERTScore for describing scholars' re⁃ search interests. Through Analytic Hierarchy Process (AHP), it was determined that our fusion model combining BERTopic and LLM outperformed other evaluation models in recognizing scholars' research interests with a manual evaluation satisfaction rate reaching 81.4%.【Innovation/limitation】The proposed model demonstrates improved capability in identifying scholars' research topics while generating high-quality descriptions of their research interests.The utilized corpus predominantly consists of Chinese texts; hence, this model exhibits limited recognition ability for foreign language papers.