plan.md

final result

• party
• party relationship, with
  • entity as party (org, company, person)
  • topic as party
  • url and archive link as links

rules

• only one relation regardless of multiple mentions
• leave possible spark for future scaling

strat

• get text from warc
• preprocess text
• get entities from text
• get topics from text - yes but expensive, maybe do on demand

__

• form party relationships
• discard articles with no entities
• save party relationships with article clean text and ref back to warc