Welcome to DatMap Platform
Research moves fast. Thousands of papers and patents appear every month, and reading lists don't scale. Data maps offer a different way in: embed every document, project the embedding space to two dimensions, cluster it hierarchically, and let an LLM name what each cluster is about. The result is a zoomable landscape of an entire field.
What you can do here
- Browse curated maps of research areas, patent families and startup ecosystems — free, no account.
- Search any map semantically once you create a free account: type a question, see the matching points light up.
- Export maps for offline use, or pull the underlying data as CSV on a paid plan.
- Build your own maps on the Pro plan: upload a CSV, or pull fresh data from Semantic Scholar and USPTO, then run the clustering pipeline in your private workspace. You'll get an email when your map is ready.
The pipeline, briefly
- Documents → sentence-transformer embeddings
- Embeddings → 2-D layout (UMAP)
- Layout → hierarchical clusters (HDBSCAN-family)
- Clusters → human-readable names (LLM)
- Everything → an interactive HTML map you can pan, zoom and query
More deep dives coming soon.