TY - GEN
T1 - Evaluation of Embedded Vectors for Lexemes and Synsets Toward Expansion of Japanese WordNet
AU - Ko, Daiki
AU - Takeuchi, Koichi
N1 - Funding Information:
A part of the research reported in this paper is supported by JSPS KAKENHI (JP19K00552) and the NINJAL project ?Development of and Research with a parsed corpus of Japanese? by JSPS KAKENHI (JP15H03210).
Publisher Copyright:
© 2020, Springer Nature Singapore Pte Ltd.
PY - 2020
Y1 - 2020
N2 - In this paper, we discuss the possibility to expand Japanese WordNet using AutoExtend that can produce embedded vectors based on dictionary structure. Recently several kinds of NLP tasks showed that the distributed representations for words are effective, however, the word-embedded vectors constructed based on contexts of surrounded words would be difficult to discriminate meanings of a word because every vector is produced for a word. On the other hand, AutoExtend that can produce embedded vectors for meanings and concepts as well as words taking into account thesaurus structure of dictionary, has been proposed and applied into English WordNet. Thus, in this paper, we apply AutoExtend into a Japanese dictionary i.e., Japanese WordNet to construct embedded vectors for lexems and synsets as well as words taking into account thesaurus structure of Japanese WordNet. The experimental results show that embedded vectors constructed by AutoExtend can be helpful to find corresponding meanings for unregistered words in the dictionary.
AB - In this paper, we discuss the possibility to expand Japanese WordNet using AutoExtend that can produce embedded vectors based on dictionary structure. Recently several kinds of NLP tasks showed that the distributed representations for words are effective, however, the word-embedded vectors constructed based on contexts of surrounded words would be difficult to discriminate meanings of a word because every vector is produced for a word. On the other hand, AutoExtend that can produce embedded vectors for meanings and concepts as well as words taking into account thesaurus structure of dictionary, has been proposed and applied into English WordNet. Thus, in this paper, we apply AutoExtend into a Japanese dictionary i.e., Japanese WordNet to construct embedded vectors for lexems and synsets as well as words taking into account thesaurus structure of Japanese WordNet. The experimental results show that embedded vectors constructed by AutoExtend can be helpful to find corresponding meanings for unregistered words in the dictionary.
KW - AutoExtend
KW - Japanese WordNet
KW - Synsets
UR - http://www.scopus.com/inward/record.url?scp=85088504691&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85088504691&partnerID=8YFLogxK
U2 - 10.1007/978-981-15-6168-9_7
DO - 10.1007/978-981-15-6168-9_7
M3 - Conference contribution
AN - SCOPUS:85088504691
SN - 9789811561672
T3 - Communications in Computer and Information Science
SP - 79
EP - 87
BT - Computational Linguistics - 16th International Conference of the Pacific Association for Computational Linguistics, PACLING 2019, Revised Selected Papers
A2 - Nguyen, Le-Minh
A2 - Tojo, Satoshi
A2 - Phan, Xuan-Hieu
A2 - Hasida, Kôiti
PB - Springer
T2 - 16th International Conference of the Pacific Association for Computational Linguistics, PACLING 2019
Y2 - 11 October 2019 through 13 October 2019
ER -