面向汉语历史事件元素识别的深度学习模型构建及实现

Construction and Implementation of Deep Learning Model for Chinese Historical Event Element Recognition

[目的/意义]命名实体识别作为文本挖掘领域的一个研究重点,是信息检索、问答系统、信息抽取等多个领域的研究基础。目前针对历史文本的研究相对较少,尤其缺乏历史文本中的命名实体识别相关研究。[方法/过程]为了较好地识别出历史事件名,笔者尝试建立深度学习模型BiLSTM-CRF进行命名实体识别。此外,基于这种模型识别了时间、地点、人物这三个辅助元素作为事件的补充信息,对不同实体的识别特性进行对比分析。[结果/结论]发现历史事件名实体识别F1值能达到95.02%。对于辅助实体元素,除地点实体F1值低于80%,其他实体元素均表现较好,F1值能达到80%以上,说明该模型具有良好的识别性及可移植性。[局限]数据标注量还不够充分,标注粒度还可以进一步细化。

[Purpose/significance] Named entity recognition,as a research focus in the field of text mining,is the research foundation in many fields such as information retrieval,question answering systems,information extraction. [Method/process] In order to better identify the names of historical events,the author attempts to establish deep learning model Bi LSTM-CRF for named entity recognition. In addition,the author still recognizes the three auxiliary elements of time,place,and person based on this model as supplementary information for the event,compares and analyzes the recognition characteristics of different entities. [Result/conclusion] It is found that the F1 value of historical event name entity recognition can reach 95. 02%. For the auxiliary entity elements,except the location entity F1 value is lower than 80%,other entity elements perform well,and the F1 value can reach more than 80%,indicating that the model in this paper has good recognizability and portability. [Limitations]The amount of data annotation is not enough,and the granularity of annotation can be further refined.