The Evolution of Themes in Artworks
In this project, I used three machine learning methods to analyze the trends in artwork themes across history.
The three approaches are:
- Word frequency analysis
- Topic modeling using BERT
- Dimensionality reduction followed by Clustering (K-Means and DBSCAN)
Each approach revealed a unique aspect of the subject matter.
I discovered that portrait, water bodies and nude bathers are the three most popular themes in art throughout history. Moreover, modern art movements has contributed to the proliferation of “untitled” artworks, making it the most frequently used artwork title across history. My analysis also revealed popular themes in specific time periods.
Please click here for more details of the project.