Contrastive Lexical Diffusion Coefficient: Quantifying the Stickiness of the Ordinary

Area of Science:

• Computational Linguistics
• Social Network Analysis
• Sociolinguistics

Background:

• Most lexical diffusion models focus on adopting new terms, neglecting changes in existing word usage.
• Lexical phenomena spread through social networks at varying rates, with existing concepts potentially diffusing more than novel ones.

Purpose of the Study:

• To introduce and evaluate a new metric, the contrastive lexical diffusion (CLD) coefficient, for quantifying the spread of common word clusters over social networks.
• To assess the degree to which ordinary language usage changes over time and across friendship connections.

Main Methods:

• Development of the contrastive lexical diffusion (CLD) coefficient to measure the diffusion of word clusters.
• Quantitative and qualitative evaluation of the CLD coefficient using six years of Twitter data.
• Analysis of tweet spread, friendship connections, and human judgments of lexical diffusion.

Main Results:

• The CLD coefficient effectively predicts tweet and friendship connection spread, correlating highly with human judgments (r=0.92) and showing replicability across networks (r=0.85).
• Topics like 'meeting' and 'job' demonstrated higher "stickiness" (diffusion) compared to negative emotions or global events.
• Positive emotion words were found to be more diffusive than negative ones, and terms like 'we' diffused more than other pronouns; numbers and time showed non-contagious patterns.

Conclusions:

• The CLD coefficient offers a robust method for understanding lexical diffusion dynamics in social networks.
• Lexical diffusion is influenced by word category, with emotional valence and social context playing significant roles.
• The study highlights the importance of analyzing changes in existing language use alongside the adoption of new terms for a comprehensive understanding of lexical spread.