Analyzing the Evolution of Social Concepts Using Temporal Embedding Models

Analyzing the Evolution of Social Concepts Using Temporal Embedding Models

Seyedeh samira Rabinataj1

1) Master s student at the University of Turin, Turin, Italy

IOI Code : AZEK-AACDFD
Publication : 4th International Conference on New Research & Achievements in Science, Engineering & Technologies(setbconf.com)
Abstract :
    The meanings of social concepts are not fixed—they evolve in response to cultural, political, and technological changes. Capturing this semantic evolution is essential for understanding shifts in public discourse and societal values. Traditional word embedding models like Word2Vec provide static representations and are thus unable to reflect temporal variations in meaning. In this paper, we introduce a dynamic embedding-based framework to trace the semantic trajectories of key social concepts—such as freedom, identity, and equality—across multiple decades. Using temporally segmented text corpora from 1950 to 2020, we train time-aware embeddings to visualize and quantify changes in meaning over time. Our results reveal significant and interpretable semantic shifts, often aligning with major historical events and social movements. This research underscores the potential of temporal embeddings as a powerful tool for computational social science and digital humanities.

Keywords : dynamic embeddings semantic shift social concepts cosine similarity temporal analysis concept evolution