Three linked views over the same corpus. Start with the document tree to read how one opinion is built and reasoned, the discourse-path map to see the corpus-wide structure of how willfulness findings connect to damage decisions, or the EDU-level map to position every case by the fine-grained rhetorical composition of its reasoning. All three are cross-linked — every case and judge opens in the document tree, and vice versa.
Document tree
Each opinion laid out as an outline of purpose-labelled sections (background, infringement, liability, relief & damages, attorney's fees, order…), with the full paragraph text and inline highlights.
- Metadata & extracted features in the left column
- Separated damage awards with source-paragraph provenance
- Inline citations & footnotes, grouped by citation type
- Click a judge, work type, or willfulness label → entity view
Discourse-path map
Every directed discourse path from a willfulness finding to a damage decision, clustered by rhetorical shape and laid out in 2-D. Colour by cluster, damage-per-work, judge experience, or work category.
- Natural-language search over paths, cases & judges
- Trace a judge → their cases → their paths
- Willfulness-status colouring & year-range filter
- Path-distribution report across the corpus
EDU-level discourse map
Every case positioned by the relation composition of its elementary discourse units — the finest rhetorical granularity in the corpus, and the representation measured to be the most informative of the three document-tree designs.
- Colour by award, willfulness, dominant relation or hierarchy depth
- Weight relations over the whole opinion or the damages sections only
- Per-case relation composition across 15 RST relation types
- Click any case → its document tree
One server, one port. Run PORT=8805 HOST=0.0.0.0 python3 app.py
in this folder, then open http://127.0.0.1:8805. Routes:
/doctree/, /map/, /edumap/.