Tracking word semantic change in biomedical literature

Area of Science:

• Computational linguistics
• Bioinformatics
• Scholarly communication analysis

Background:

• Research on scholarly communication has predominantly focused on syntax, neglecting semantic shifts.
• Understanding semantic change within specific scientific disciplines like biomedicine remains underexplored.

Purpose of the Study:

• To analyze and illustrate word semantic change in biomedical literature.
• To quantify semantic stability at both word and topic levels over time.

Main Methods:

• Identification of representative biomedical terms using word frequency and topic probability.
• Application of a word2vec language model to measure semantic changes.
• Analysis of global and local semantic distances for topics and words.

Main Results:

• Biomedical word meanings show increased stability in the 2000s compared to the 1980s and 1990s.
• Most topics (19/20) exhibit declining global and local distances, indicating semantic stabilization of associated terms.
• Two word-level semantic trends observed: semantic clustering and co-evolution, or divergence from neighbors while stabilizing globally.

Conclusions:

• Biomedical discourse is experiencing semantic stabilization, contrary to some linguistic theories.
• Findings suggest a complex interplay of semantic convergence and divergence within scientific language.
• The study provides quantitative evidence for semantic shifts in a major scientific field.