[Paper Review] Characterizing Financial Market Coverage using Artificial Intelligence

Jean Marie Tshimula, D’Jeff K. Nkashama+8 more|arXiv (Cornell University)|Feb 7, 2023

FinTech, Crowdfunding, Digital FinanceBusiness, Management and Accounting3 citations

TL;DR

This paper uses OpenAI's Whisper to transcribe over 4,900 YouTube financial market videos from Bloomberg and Yahoo Finance, applying NLP techniques like topic modeling and named entity recognition to analyze language use, trending topics, and media coordination. It reveals that financial news coverage centers on major global events—such as the Russia-Ukraine war and pandemic-related streaming surges—and shows strong content coordination across channels on key topics like inflation, recession, and mortgage rates.

ABSTRACT

This paper scrutinizes a database of over 4900 YouTube videos to characterize financial market coverage. Financial market coverage generates a large number of videos. Therefore, watching these videos to derive actionable insights could be challenging and complex. In this paper, we leverage Whisper, a speech-to-text model from OpenAI, to generate a text corpus of market coverage videos from Bloomberg and Yahoo Finance. We employ natural language processing to extract insights regarding language use from the market coverage. Moreover, we examine the prominent presence of trending topics and their evolution over time, and the impacts that some individuals and organizations have on the financial market. Our characterization highlights the dynamics of the financial market coverage and provides valuable insights reflecting broad discussions regarding recent financial events and the world economy.

Motivation & Objective

• To address the challenge of extracting actionable insights from the overwhelming volume of financial market videos on YouTube.
• To investigate how language use, trending topics, and key entities (individuals and organizations) shape financial news narratives across platforms.
• To assess content coordination in financial news coverage across major media channels like Bloomberg and Yahoo Finance.
• To provide a reliable, open-source framework for analyzing video-based financial discourse using speech-to-text and NLP pipelines.

Proposed method

• Employed OpenAI’s Whisper model to transcribe audio from 4,900 YouTube financial market coverage videos into text.
• Applied topic modeling to identify and track evolving themes in financial and economic discourse over time.
• Conducted n-gram analysis and named entity recognition (NER) to extract frequently mentioned persons, organizations, and key phrases.
• Focused analysis on topics related to the economy and financial markets, excluding non-relevant content.
• Compared narrative structures and topic similarities across different media channels to detect coordination in coverage.
• Released code and data publicly to support reproducibility and further research in financial media analysis.

(a) Bloomberg Wall Street Week (BLW)

Experimental results

Research questions

• RQ1 How are major financial events identified through language use within news coverage topics?
• RQ2 To what extent do news coverage topics exhibit content coordination regarding major financial events and entities (such as organizations and individuals) across different news channels?
• RQ3 How do the frequencies and evolutions of key topics (e.g., inflation, recession, Russia-Ukraine war) correlate with real-world financial and economic developments?
• RQ4 Which individuals and organizations are most frequently cited in financial market coverage, and how does their mention frequency relate to market events?

Key findings

• The topic 'inflation' saw a significant increase in frequency from 2021 to 2022, closely linked to rising mortgage rates and recession concerns.
• The 'Russia and Ukraine' topic surged in early 2022, correlating with the onset of the Russo-Ukrainian war and global economic sanctions.
• The 'Disney and Netflix content' topic experienced a sharp rise during the first COVID-19 lockdown, reflecting increased streaming consumption and stock performance.
• Financial news coverage exhibits strong coordination across channels, particularly on high-impact events, with consistent use of bi-grams and named entities.
• Named entity recognition identified prominent figures and institutions as central to market narratives, often tied to economic policy or market-moving developments.
• Topic modeling revealed that coverage is predominantly reactive, with narratives shifting rapidly in response to global events and market volatility.

(b) Bloomberg Stock Market News and Analysis (BSM)