Using meaning instead of words to track topics

TL;DR

This paper explores a novel semantic-based approach for tracking topics over time using word embeddings, demonstrating comparable performance to traditional lexical methods and suggesting potential for complementary use.

Contribution

Introduces a semantic-based topic tracking method using word embeddings, filling a gap in existing lexical-only approaches.

Findings

Semantic approach performs on par with lexical methods

Semantic and lexical methods make different errors

Both methods could be combined for improved tracking

Abstract

The ability to monitor the evolution of topics over time is extremely valuable for businesses. Currently, all existing topic tracking methods use lexical information by matching word usage. However, no studies has ever experimented with the use of semantic information for tracking topics. Hence, we explore a novel semantic-based method using word embeddings. Our results show that a semantic-based approach to topic tracking is on par with the lexical approach but makes different mistakes. This suggest that both methods may complement each other.