Paid DailyPDPaid Daily
Open the desk
QuantInsti Quantitative Learning

How Hedge Funds Trade News Data: What NLP Can Really Extract From Text

NowPress play to follow along0:00 / 6:58
Chapters3 segments · tap to seek
ticker lesson

Breakdown

0:00lesson

News Data Formats and Metadata — JSON/XML, programmatic consumption

  • News data formats include JSON/XML for programmatic reading
  • Metadata includes timestamp, headline, language, source, subjects
  • NLP enables structured sentiment analysis for filtering news
  • Three-class sentiment scores: positive/negative/neutral
2:02lesson

Sentiment Probabilities — 3-class scores for companies

  • Sentiment scores: +1/-1/0 based on highest probability class
  • Hedge funds use precomputed sentiment signals directly
  • Micro-sentiments include volatility, production volume, urgency
  • News items tagged with entities: companies, people, orgs
4:02lesson

Kalman Filter for Noise Reduction — State Space Models

  • Kalman filter reduces noise in high-frequency sentiment data
  • State space models process text-based signals for trading
  • Granger causality tests sentiment's predictive power
  • Entity relationships: affiliates, acquisitions, joint ventures

Informational only — this is QuantInsti Quantitative Learning’s content, decoded by Plutus. Not Paid Daily’s advice or a recommendation. The outline, timestamps, and claims are extracted from what the creator said; verify before acting.