Domain-General Versus Domain-Specific Named Entity Recognition: A Case Study Using TEXT

Abstract

Named entity recognition (NER) seeks to identify and classify named entities within bodies of text into language categories such as nouns, that are reflective of locations, organizations, and people. As it is language dependent, the approach taken for most NER systems are domain-general, meaning that they are designed based on a language and not on a specific targeted domain. With current usage of non-formal languages on social media, this instigates the need to compare the performance of domain-general and domain specific NERs. A domain specific NER (vehicle traffic domain), TEXT, is described and the performance of domain-general NER versus TEXT is compared. The results of the evaluation show that the performance of domain-specific NER significantly outperforms domain-general NER. The domain-general NER could only perform adequately for common scenarios.