Using text analysis to quantify the similarity and evolution of scientific disciplines

Area of Science:

• Bibliometrics
• Scientometrics
• Information Science

Background:

• Understanding the structure and evolution of scientific fields is crucial for research policy and scientific progress.
• Traditional methods like expert opinion and citation analysis offer valuable insights but have limitations.

Purpose of the Study:

• To investigate the organization and evolution of scientific fields using an information-theoretic measure of linguistic similarity.
• To compare linguistic similarity with expert and citation-based classifications for a more comprehensive view of science.

Main Methods:

• Analysis of nearly 20 million scientific papers published over the past three decades.
• Application of an information-theoretic measure to quantify linguistic similarity between scientific fields.
• Temporal analysis to track changes in field similarity over time.

Main Results:

• Linguistic similarity provides a complementary perspective to expert and citation-based classifications of science.
• Certain fields, such as computer science, are increasing in centrality within the scientific landscape.
• The average similarity between pairs of scientific disciplines has remained relatively stable over recent decades.

Conclusions:

• Linguistic similarity offers a novel and valuable approach to mapping the structure of science.
• The balance between disciplinary convergence and divergence suggests a dynamic yet stable scientific ecosystem.
• Future research can leverage linguistic similarity for enhanced understanding of scientific trends and interdisciplinary connections.