CTLKE: Keyphrase Extraction Agent for Chinese Technical Literature

Keyphrase extraction is a fundamental task in natural language processing and has been widely applied in areas such as text classification, information retrieval, and technology trend prediction. However, most existing studies primarily focus on English corpora, while high-quality Chinese keyphrase datasets remain scarce. To address this limitation, this paper constructs a Chinese technical literature dataset in emerging technology domains and proposes an unsupervised keyphrase extraction algorithm CTLKE based on pre-trained language models. The proposed algorithm leverages the contextual semantic under-standing capability of pre-trained models to enhance the semantic representation of candidate phrases, thereby improving the accuracy of keyphrase extraction. Comparative experiments with classical unsupervised methods, including TF-IDF, RAKE, YAKE, TextRank, and KeyBERT, demonstrate that CTLKE achieves superior performance, with a Macro F1-score of 0.5099 and a Micro F1-score of 0.5073, outperforming all baseline algorithms. Experimental results indicate that CTLKE can effectively extract high-quality keyphrases from Chinese technical literature, and that domestically developed open-source language models exhibit better adaptability to Chinese datasets.