Named entity recognition for corporate names in news text data and identification of the name by characteristics of neighboring words

Abstract

We used huge news data distributed by SPEEDA service, which include not only news on general interest but also business and industry-specific topics, and built a model to extract corporate information appearing in the text as named entity. In this study, we proposed a method to extract the corporate names in the text after segmented by tokenizers and the extracted were matched with a corporate name dictionary added with their automatically-generated abbreviations and so on. Thereby, we succeed in extracting a named entity which is identified as a corporate name and the method improved the accuracy of the task of extracting corporate names and identification of the company.