Named Entity Recognition Method for Document-Level Label Consistency

Abstract

When trying to recognize named entity for the purpose of anonymization of personal information and information extraction in documents, statistical approaches such as machine learning may be given different labels for the similar expressions. If the same label is desirable to consistently impart, by realizing the determination method, it is possible to improve the accuracy of labeling. This study proposes a method to determine the consistency of such named entities. In the experiment using the Extended named entity annotated corpora, the improvement of accuracy was confirmed from the baseline.