Semantic coordinates analysis reveals language changes in the AI field
This work addresses the challenge of tracking semantic evolution in rapidly changing academic communities, offering a tool for analyzing trends, though it is incremental as it builds on existing semantic shift methods.
The authors tackled the problem of detecting language changes in fast-evolving fields like AI over short time spans, proposing semantic coordinates analysis to reveal shifts in research interests and activities, with results showing detectable changes in terms like 'deep' and 'collaboration' within as little as 10 years.
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.