[4G1-GS-6] Language media processing:

[4G1-GS-6-05] Similarity Extraction in Time-series Data Using Natural Language Transformation

〇Mahiro Tajima¹, Teruaki Hayashi¹ (1. Univ. of Tokyo)

Keywords:Natural language processing, Time series analysis, Information retrieval, Verbal representation

Statistical methods and pattern recognition techniques often fail to capture local, time-lagged similarities in time-series data, whereas humans intuitively recognise such patterns. This study proposes an approach by converting time-series data into natural language representations. The data are segmented into short sequences, transformed into natural language expressions, and processed using NLP techniques to compute sequence similarities. Results show that this approach effectively detects local similarities across different time-series data.