Why splitting legal documents for AI can be difficult

A personal project uses legal and government documents. Early has found more problems in the individual text chunks than in the system as a whole.

The central question is whether is generally difficult or especially difficult for legal material. No specific splitting rules, test results, or verified solution are provided.

Key points

  • The project works with legal and government documents.
  • Most problems are appearing at the chunk level.
  • It is unclear whether the difficulty comes from itself or the legal domain.
  • No tested method or cost-saving result is included.
Read original