News Headline Analysis

What follows is an analysis of Internet news headlines. To create the dataset, a script ran daily to capture the headlines from the news sites that follow:

• ABC: https://abcnews.go.com/
• AP: https://apnews.com/
• Breitbart: https://www.breitbart.com
• CBS: https://www.cbsnews.com/
• CNN: http://rss.cnn.com/rss/cnn_topstories.rss
• Fox: https://www.foxnews.com/
• New York Post: https://nypost.com/
• NBC: http://feeds.nbcnews.com/nbcnews/public/news
• NPR: https://www.npr.org
• OAN:https://www.oann.com
• Reuters: https://www.reuters.com/
• Washington Post: https://www.washingtonpost.com/

NBC and CNN use RSS feeds. The rest are those pages are built as the site's home page.

This analysis currently includes headlines from July 5 to August 2, 2021. The exception is OAN, which was brought into the analysis on July 10.

Average Compount Sentiment Analysis

This analysis looks at the overall compound sentiment for the period of review. Sentiment values range from -1 (negative sentiment) to 1 (positive sentiment) with 0 as neutral sentiment.

The table and graph below organizes the source by most to least positive sentiment.

        source_name  compound
0           Reuters  0.012618
1               CNN -0.007112
2   Washington Post -0.017253
3               NPR -0.031112
4               Fox -0.043821
5               CBS -0.050776
6               OAN -0.051326
7     New York Post -0.067105
8                AP -0.076693
9               ABC -0.090861
10        Breitbart -0.092640
11              NBC -0.120787

Compound Sentiment Boxplots

What follows are boxplots for the compound sentiment over the same period. This provides more nuance into the range of values than an average alone provides.

For example, NPR has most of its range in tight focus with many outliers. This is in contrast to Fox's wide range of values. CNNs relatively even distribution is a point of curiosity as well.

Compound Sentiment by Date

The Compound Sentiment by Date allows us to see how news headline sentiment trends on each site or feed. This analysis also provides an Overall average for comparison. Select or deselect sites to have them appear or hide in the graph below.

Points of interest include:

• CNN and the Washington Post have average trends higher than the Overall average.
• Fox's average trends close to the Overall average.
• NPR's average starts close to the Overall average, but then trends higher at the start of July 19.
• New York Posts's average is close to the Overall average, with a noticble dip between July 23 and 26.
• OAN crosses the average in an pattern of interest.