CNN International Tweets Sentiment Analysis

Analyzing CNN International's Tweets to determine the underlying sentiments and subjectivity.

View Project File

Summary of Data Analysis

Dataset

Sentiment Analysis of a dataset of 200 Tweets from CNN International collected through the Twitter API on October 10th, 2021.

• The data indicates tweet id, source, source_url, user, the full text of the tweet, number of favorites, and number of retweets. There are no missing values in the data. 

• In addition to Sentiment Analysis, the relationships with the text content of the tweets, the number of favorites, and the number of retweets are also examined.

Tools

Python, Twitter API

Insights

The keyword ‘US’ appears most often on CNN International’s Tweets, and it is a top keyword in all types of Tweets: Positive, Negative and Neutral. There was also some degree of overlap between the top Keywords in both Positive and Negative Tweets, suggesting that some Tweets may reflect both sides of a news article.

• Positive Tweets represent the biggest portion of Tweets, followed by Neutral and then Negative Tweets.

• However, Negative Tweets receive the most favorites and retweets on average, which suggests that Negative Tweets are the most popular despite appearing the least often.

• The majority of CNN International’s popular tweets, and their tweets in general, tend to have a Subjectivity of less than 0.5 and a Polarity of between -0.25 to +0.25. The observation suggests that most Tweets are close to Neutral. As CNN International is a news source this sentiment would make sense as it would be the result of an intentional attempt to post relatively neutral, informative content, allowing the readers to make their own judgment.

Limitations

The Analysis revealed that there are clearly some tweets that are placed in the wrong sentiment categories. The potential causes for this could be as follows:

• The range for Positive, Negative, and Neutral Sentiments needs to be adjusted to more accurately reflect these sentiments.

• The Polarity may not provide an accurate assessment of the Tweet Sentiment; a different metric may be necessary.

• A different function might be necessary to clean the data prior to calculating Polarity and Subjectivity.