joaoli131

5 ms·

1. 1.▲
   by joaoli131 6mo ago

   Adding a temporal dimension is a great idea. It would offer a fascinating perspective on the data by allowing us to track the 'semantic drift' of legal frameworks over time. Brazil is actually a perfect use case for this: the coun
2. 2.▲
   Show HN: Semantic atlas of 188 constitutions in 3D (30k articles, embeddings)(constitutionalmap.ai)

   12 points by joaoli131 6mo ago|4 comments
3. 3.▲
   by joaoli131 1y ago

   Embeddings are still underrated—even in RAG. Legal text is deeply hierarchical and full of pointers (“Art. 5 CF”, “see Art 34”). One vector per article leaves too much on the table. Things that moved the needle for us: – *Multi-layer embeds