A novel embedding approach to learn word vectors by weighting semantic relations: SemSpace
作者:
Highlights:
• SemSpace is a novel -embedding - approach to determine words vectors in a lexical semantic network.
• The idea behind the proposed approach is the learning vector quantization method.
• It utilizes a dual optimization technique to obtain semantic relation weights and word vectors.
• Remarkable benchmark results are obtained in word-level semantic similarity tasks.
摘要
•SemSpace is a novel -embedding - approach to determine words vectors in a lexical semantic network.•The idea behind the proposed approach is the learning vector quantization method.•It utilizes a dual optimization technique to obtain semantic relation weights and word vectors.•Remarkable benchmark results are obtained in word-level semantic similarity tasks.
论文关键词:SemSpace,Embedding,Word vectors,Aligning semantic relations to weights,WordNet
论文评审过程:Received 31 January 2020, Revised 20 March 2021, Accepted 29 April 2021, Available online 4 May 2021, Version of Record 10 May 2021.
论文官网地址:https://doi.org/10.1016/j.eswa.2021.115146