charieswei
/
TradingAgents
mirror of https://github.com/TauricResearch/TradingAgents.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
TradingAgents/docs/iterations/scanners/semantic_news.md

595 B
Raw Blame History

Semantic News Scanner

Current Understanding

Currently regex-based extraction, not semantic. Headline text is not included in candidate context — the context just says "Mentioned in recent market news" which is not informative. Catalyst classification from headline keywords (upgrade/FDA/ acquisition/earnings) would improve LLM scoring quality significantly.

Evidence Log

(populated by /iterate runs)

Pending Hypotheses

  • Would embedding-based semantic matching outperform keyword regex?
  • Does catalyst classification (FDA vs earnings vs acquisition) affect hit rate?