Named Entity Recognition by Character-based Word Classification using a Domain Specific Dictionary

EasyChair Preprint 1728

Abstract

Named entity recognition is a fundamental task in natural language processing and has been widely studied. The construction of a recognizer requires training data that contains annotated named entities. However, it is expensive to construct such training data for low-resource domains. In this paper, we propose a recognizer that uses not only training data but also a domain specific dictionary that is available and easy to use. Our recognizer first uses character-based distributed representations to classify words into categories in the dictionary. The recognizer then uses the output of the classification as an additional feature. We conducted experiments to recognize named entities in recipe text and report the results to demonstrate the performance of our method.

Keyphrases: Named Entity Recognition, neural network, recipe text

@booklet{EasyChair:1728,
  author    = {Makoto Hiramatsu and Kei Wakabayashi and Jun Harashima},
  title     = {Named Entity Recognition by Character-based Word Classification using a Domain Specific Dictionary},
  howpublished = {EasyChair Preprint 1728},
  year      = {EasyChair, 2019}}