본문 바로가기
  • Home

The word embedding and the semantic relation

  • Korean Semantics
  • 2020, 70(), pp.1~49
  • DOI : 10.19033/sks.2020.12.70.1
  • Publisher : The Society Of Korean Semantics
  • Research Area : Humanities > Korean Language and Literature
  • Received : August 18, 2020
  • Accepted : December 17, 2020
  • Published : December 30, 2020

KIM SO HEE 1 Do Jaehak 2

1인천대학교 코퍼스연구소
2경기대학교

Accredited

ABSTRACT

This paper investigates semantic relation which is captured by the word embedding(Word2vec). As precedent researches using this methodology already have reviewed the effectiveness of study on the semantic similarity or relativeness of lexical units, it is focused on that which specific properties could be identified. In particular, not only lexical units but also grammatical units were included in this study as the objects. To sum up, it is hard to say that a certain specific type of relation such as synonymy, antonymy, hyponymy, meronymy, or contrastive relation is caught saliently by the word embedding. Rather various types of relations are widely observed since the related words which compose lexical field with target word are extracted under the influence of individual context(especially register). Generally, the results are heavily dependent on the contexts of word usage. Meanwhile, even though it was not enough in terms of grammatical units, there are two small but significant achievements. The first is that the word embedding shows relatively better results in some grammatical categories. And the second is that periphrastic constructions are worthy to be considered equal to single morpheme concerning grammatical function.

Citation status

* References for papers published after 2023 are currently being built.