面向态靶辨治的语料库标注系统设计与实现

Design and Implementation of a Corpus Annotation System for State-Target Differentiation and Treatment

摘要: 针对中医态靶辨治领域非结构化文本阻碍知识挖掘的问题，本研究提出一种面向对象的语料库标注系统解决方案。基于筛选自CNKI的32篇态靶辨治核心文献，系统化定义10类实体(如疾病、标靶、症型态)与7类实体关系(如打靶、靶药、药理)，并设计三层次标注框架：实体标注、实体属性标注、实体关系标注。系统实现了可视化的标注界面与BIE结构化的输出，支持人工高效构建高质量标注语料库。本工作既可以为后续基于深度学习的态靶辨治文本标注提供训练数据集，又能为中医同类标注系统研发提供参考。

Abstract: To address the challenge of unstructured text hindering knowledge mining in the domain of State-Target Differentiation and Treatment (STDT) of Traditional Chinese Medicine, this study proposes an object-oriented corpus annotation system solution. Based on 32 core STDT literature pieces curated from CNKI, we systematically define 10 entity categories (e.g., disease, clinical indicator, syndrome state) and 7 entity relationships (e.g., precision targeting, target-specific herb, pharmacological mechanism). A three-tier annotation framework is designed: entity annotation, entity attribute annotation, and entity relationship annotation. The system implements a visualized annotation interface with BIE-structured output, enabling efficient manual construction of high-quality annotated corpora. It not only provides a training dataset for deep learning-based STDT text annotation, but also serves as a reference framework for developing analogous TCM annotation systems.