Abstract
Natural language processing aims to allow computers to understand human language. One of the main challenges for computers in understanding natural language is the ambiguity inherent in human language. Recently, language models using transformers have shown excellent performance in natural language processing. In this paper, we propose a semantic vector visualization method that enables the high-dimensional embedding space of the transformer-based language model to be projected onto a two-dimensional visualization space through clustering and semantic analysis. We use Word2vec as a word embedding technology. We demonstrate the practical use of the proposed visualization method for analyzing the similarity and relatedness of word vectors through a case study. The proposed method enables an intuitive understanding of the semantic relationships between words and can be used for various natural language understanding tasks.