Semantic coordinates analysis reveals language changes in the AI field

TL;DR

Semantic coordinates analysis is a novel method that detects language and research trend changes in rapidly evolving scientific communities like AI over short periods, revealing shifts in research focus and collaboration patterns.

Contribution

The paper introduces semantic coordinates analysis, a new technique for identifying language and research trend shifts in short-term scientific publication data.

Findings

Detects research interest shifts, e.g., 'deep' from 'rigorous' to 'neural'

Identifies changes in research activities, e.g., 'collaboration' less associated with 'competition'

Works on publication data spanning as short as 10 years

Abstract

Semantic shifts can reflect changes in beliefs across hundreds of years, but it is less clear whether trends in fast-changing communities across a short time can be detected. We propose semantic coordinates analysis, a method based on semantic shifts, that reveals changes in language within publications of a field (we use AI as example) across a short time span. We use GloVe-style probability ratios to quantify the shifting directions and extents from multiple viewpoints. We show that semantic coordinates analysis can detect shifts echoing changes of research interests (e.g., "deep" shifted further from "rigorous" to "neural"), and developments of research activities (e,g., "collaboration" contains less "competition" than "collaboration"), based on publications spanning as short as 10 years.