基于类语言模型的中文机构名称自动识别

提出了一种基于类语言模型的中文机构名称自动识别方法,将分词和机构名称自动识别有机地结合起来。在机构名称识别的类语言模型中采用等级结构,使得嵌套有人名、地名等实体的机构名称能够较好地识别出来。在实验过程中,逐步增加实验条件,依次加入启发信息、缓存模型和机构名缩写处理,使得实验结果显著提高。在开放测试中,中文机构名称最终识别的查准率和查全率分别为85.47%和72.81%。

Chinese Organization Name Automatic Recognition Using Class-based Language Model

An approach based on class language model is put forward about Chinese organization automatic recognition. Word segmentation and organization recognition can be combined. We adopted a hierarchical structure in organization model so that the organization names including nested entities can be identified. In the experiments, we integrate heuristic information, cache model and organization abbreviation processing into the class-based language model. The precision and recall of Chinese organization recognition are 85.47 and 72.81 in the opened-test.