Paper: arXiv 2304.05115

Authors: Jianfei Zhang, Mathieu Rosenbaum

Abstract

News can convey bearish or bullish views on financial assets. Institutional investors need to evaluate automatically the implied news sentiment based on textual data. Given the huge amount of news articles published each day, most of which are neutral, we present a systematic news screening method to identify the true'' impactful ones, aiming for more effective development of news sentiment learning methods. Based on several liquidity-driven variables, including volatility, turnover, bid-ask spread, and book size, we associate each 5-min time bin to one of two specific liquidity modes. One represents the calm’’ state at which the market stays for most of the time and the other, featured with relatively higher levels of volatility and trading volume, describes the regime driven by some exogenous events. Then we focus on the moments where the liquidity mode switches from the former to the latter and consider the news articles published nearby impactful. We apply naive Bayes on these filtered samples for news sentiment classification as an illustrative example. We show that the screened dataset leads to more effective feature capturing and thus superior performance on short-term asset return prediction compared to the original dataset.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper introduces a practical, liquidity-based regime-switching framework to filter impactful news for sentiment modeling. While the mathematical modeling relies on basic classification techniques like Naive Bayes, the empirical approach addresses a real-world market microstructure problem with clear predictive utility.

Research Flowchart

  flowchart TD
    A["Research Goal: Identify impactful intraday news<br>to improve sentiment learning & return prediction"] --> B["Input Data: Financial news articles &<br>5-min liquidity metrics (volatility, turnover, spread, book size)"]
    B --> C["Liquidity Mode Screening: Classify 5-min bins<br>into Calm vs. Event-driven modes"]
    C --> D["Filter News: Select articles published near<br>Calm-to-Event regime switches"]
    D --> E["Sentiment Classification: Train Naive Bayes model<br>on filtered impactful news"]
    E --> F["Outcomes: Enhanced feature capturing &<br>superior short-term asset return prediction"]