Korean named entity recognition based on language-specific features

Abstract In this paper, we propose a novel way of improving named entity recognition (NER) in the Korean language using its language-specific features. While field NER has been studied extensively recent years, mechanism efficiently recognizing entities (NEs) hardly explored. This is because distinct linguistic properties that present challenges for modeling. Therefore, an annotation scheme corpora by adopting CoNLL-U format, which decomposes words into morphemes and reduces ambiguity NEs original segmentation may contain functional such as postpositions particles, proposed herein. We investigate how NE tags are best represented morpheme-based implement algorithm to convert word-based syllable-based with format. Analyses results traditional neural models reveal format feasible, varied performances under influence various additional features demonstrated. Extrinsic conditions were also considered observe variance models, given different types data, including tagging formats.