Extraction of Chinese Medical Entity Attribute Values Based on Multi-type Feature Fusion

Extracting entity attribute values from Chinese electronic medical records remains a challenging issue in the field of entity recognition. The characteristics of the Chinese language, combined with the style of medical record writing, mean that the extraction of entity attributes requires more textual feature information; otherwise, issues may arise such as vague word boundaries and inaccurate entity and attribute extraction. This paper proposes a Chinese medical entity attribute value extraction model based on multi type feature fusion for Chinese electronic medical record texts. The model captures character-level, word-level, and word-category information from medical texts, enriching feature extraction and strengthening word boundary features and providing ample semantic representation. At the same time, the model utilizes a gating mechanism to realize the adaptive dynamic fusion of character contextual information and local character information within sentences, thereby highlighting character feature values in line with the language environment of the training texts, which is conducive to the formation of semantic information for entities and attribute values. The experimental results show that the model designed in this paper achieves F1 scores of 95.53% and 96.35% on the MSRA and a self-built dataset, respectively, proving the effectiveness of the model in extracting entity attribute values from Chinese electronic medical records.