CLaSP: Learning Concepts for Time-Series Signals from Natural Language Supervision

TL;DR

CLaSP is a new model that uses contrastive learning and large language models to retrieve time-series signals based on natural language descriptions, improving scalability and flexibility over previous methods.

Contribution

It introduces a contrastive learning approach that leverages LLMs to map signals to natural language, removing the need for manual synonym dictionaries and domain-specific design.

Findings

High accuracy in retrieving signals with natural language queries

Effective on TRUCE and SUSHI datasets

Outperforms existing retrieval methods

Abstract

This paper presents CLaSP, a novel model for retrieving time-series signals using natural language queries that describe signal characteristics. The ability to search time-series signals based on descriptive queries is essential in domains such as industrial diagnostics, where data scientists often need to find signals with specific characteristics. However, existing methods rely on sketch-based inputs, predefined synonym dictionaries, or domain-specific manual designs, limiting their scalability and adaptability. CLaSP addresses these challenges by employing contrastive learning to map time-series signals to natural language descriptions. Unlike prior approaches, it eliminates the need for predefined synonym dictionaries and leverages the rich contextual knowledge of large language models (LLMs). Using the TRUCE and SUSHI datasets, which pair time-series signals with natural language descriptions, we demonstrate that CLaSP achieves high accuracy in retrieving a variety of time series patterns based on natural language queries.