[Paper Review] Analyzing Who and What Appears in a Decade of US Cable TV News

James Won‐Ki Hong, Will Crichton+10 more|arXiv (Cornell University)|Aug 13, 2020

Authorship Attribution and Profiling4 references4 citations

TL;DR

This paper analyzes a decade (2010–2019) of U.S. cable TV news from CNN, FOX, and MSNBC using machine learning to detect faces, label gender presentation, and align captions with audio. It reveals a persistent gender imbalance, with male-presenting individuals receiving 1.9x more screen time than female-presenting individuals by 2019, and introduces an open, interactive web tool for public exploration of the data set.

ABSTRACT

Cable TV news reaches millions of U.S. households each day, meaning that decisions about who appears on the news and what stories get covered can profoundly influence public opinion and discourse. We analyze a data set of nearly 24/7 video, audio, and text captions from three U.S. cable TV networks (CNN, FOX, and MSNBC) from January 2010 to July 2019. Using machine learning tools, we detect faces in 244,038 hours of video, label each face's presented gender, identify prominent public figures, and align text captions to audio. We use these labels to perform screen time and word frequency analyses. For example, we find that overall, much more screen time is given to male-presenting individuals than to female-presenting individuals (2.4x in 2010 and 1.9x in 2019). We present an interactive web-based tool, accessible at https://tvnews.stanford.edu, that allows the general public to perform their own analyses on the full cable TV news data set.

Motivation & Objective

• To understand who appears on U.S. cable TV news and how representation has changed over time.
• To quantify gender disparities in screen time and identify patterns in the visibility of public figures.
• To develop a scalable, automated pipeline for labeling faces, gender presentation, and captions in large-scale broadcast video.
• To create an accessible, interactive web platform enabling public and journalistic exploration of cable news media trends.
• To provide a foundation for longitudinal, data-driven analysis of media representation and discourse in American news.

Proposed method

• Applied MTCNN face detection on video frames sampled every three seconds to identify faces across 244,038 hours of broadcast footage.
• Used machine learning models to classify the presented gender of detected faces and identify prominent public figures via facial recognition.
• Aligned text captions with audio transcripts using automatic speech recognition and temporal synchronization techniques.
• Filtered the data to focus on news programming (72.1% of total video), excluding commercials, for analysis of content and representation.
• Built a web-based interactive visualization tool that allows users to query screen time and word frequency trends over time.
• Validated labeling accuracy through manual inspection and used time-series analysis to reveal longitudinal trends in media coverage.

Experimental results

Research questions

• RQ1 How has the screen time of male- versus female-presenting individuals changed across U.S. cable news networks from 2010 to 2019?
• RQ2 Which public figures and political figures receive the most on-screen visibility in cable news?
• RQ3 How has the proportion of time with multiple faces on screen evolved over the past decade?
• RQ4 What topics are most frequently discussed, and how do they correlate with on-screen representation?
• RQ5 How does the visibility of victims and perpetrators in news coverage vary by gender and context?

Key findings

• The ratio of screen time between female-presenting and male-presenting individuals increased from 0.41:1 in 2010 to 0.54:1 in 2019, indicating a gradual improvement but persistent gender imbalance.
• By 2019, male-presenting individuals received 1.9 times more screen time than female-presenting individuals, down from 2.4x in 2010.
• The percentage of time with at least one face on screen rose from 72.9% in 2010 to 81.5% in 2019, indicating a growing emphasis on on-screen human presence.
• On CNN and FOX, the proportion of time with only one face on screen declined, while it remained stable on MSNBC.
• The top 100 individuals by screen time included 18 U.S. politicians and 85 news presenters, with 3 being both.
• The data set contains 263 million detected faces, with 75.3% of total airtime featuring at least one face on screen.