This project is an experimental visualization I carried out during university (around 4 years ago).
The goal was to explore whether programming languages and technologies could be arranged on a map, such that languages used in similar contexts appear closer together.
- Collected a large dataset of Stack Overflow posts (tags and post content).
- Represented the data in multiple ways to capture relationships between languages.
- Applied UMAP (Uniform Manifold Approximation and Projection), t-SNE (t-distributed Stochastic Neighbor Embedding), and K-Means clustering for dimensionality reduction and clustering.
- Visualized the resulting embeddings/clusters in 2D to create a "map of programming."
- Tested several approaches to represent the data.
- Ran multiple experiments using UMAP, t-SNE and K-Means, tuning hyperparameters extensively.
- Found UMAP to produce the best results.
- Investigated whether related languages (e.g., C, C++, and C#) or domains (e.g., web vs. systems programming) would cluster together.
Some of the results revealed several interesting patterns:
- In some experiments, languages used in similar ways often appeared closer together.
- Certain domains (e.g., web-related tags) tended to form clusters.
- Some languages stood out as isolated due to their more niche or domain-specific usage.
- The experiment highlighted both the promise and the challenges of trying to "map" programming ecosystems from community data.
- This was a fun side project, not a polished research pipeline.
- The code and exact preprocessing details were sadly not preserved.
- What remains is this visualization, which captures the essence of the idea.