基于数据增强与多特征组合的中文命名实体识别

打开文本图片集

中图分类号：TP391 文献标识码：A 文章编号：2096-4706（2025）16-0050-07

Chinese Named Entity Recognition Based on Data Augmentation and Multifeature Combination

LI Yuan

（School of InformationEngineering，Xinyang AgricultureandForestry University，Xinyang 4640oo，China）

Abstract：NamedEntityRecognition（NER）isanimportantand fundamentaltask inthefieldsofInformationRetrieval and Natural Language Procesing.At present，the mainstream methods based on character combination Attention Mechanism （AM）andcharacterand word combination AMare faced with problems such as corpus，Chinese word segmentationandoverfow Words.Therefore，from the perspectiveof dataset andthecombination ofcharacter and word，this paper proposes a method combining Data Augmentation （DA）and dynamic feature combinationofcharacter and worddomain information.The useof DAtechnologyimproves thequalityandexpands thescaleofcorpus，whilethedynamiccharacterand word featurescombined with domain information byusing AM provide efective textual semantic information.The paper conducts alarge numberof experiments onCCKS2o17and Commondatasets，andthe experimentalresultsshowthe effectivenessof the proposed model.

eywords：data augmentation; dynamic feature combination;Atention Mechanism; Chinese Named Entity Recognit

0 引言

作为信息检索和自然语言处理（Natural LanguageProcessing，NLP）领域重要且基础的前置任务，命名实体识别（Named Entity Recognition，NER）有着广泛的应用前景，如文献检索、病历抽取，知识图谱等。（剩余10073字）