| By Dustin Fontaine |
Social media plays a huge role in modern day politics. Through social media, politicians can instantly share their ideas and opinions with millions of people around the world. Social media posts do not have to pass through a middleman (i.e. a journalist or news-anchor…potential sources of bias), before reaching their intended audience, so many politicians use social media as one of their main platforms for communication. We all know that President Donald Trump uses Twitter heavily. What about US Senators?
It turns out that all 100 current US Senators are active on Twitter. I wanted to investigate and compare each Senator’s tweeting behavior, so I used some analytics tools, Python and Tableau, to do so. Check out my findings in the dashboard below.
This Tableau dashboard can also be viewed here.
The sections below describe how I completed this project. It is not a step by step guide, but can serve as a basic roadmap for someone wanting to complete similar work.
Gathering Tweets
The first, most tedious step was to identify the twitter handles of all 100 senators. This link provided a great start, but to ensure accuracy I manually inspected each twitter page. For senators that had multiple accounts, I decided to use the account that appeared the most “official” and the least personal.
Twitter has a nifty REST API that allows programmatic reading and writing of Twitter data. Getting an API key was simple. All I had to do was head to apps.twitter.com, sign into my Twitter account, and click the “Create New App” button. After filling out a form, checking a few boxes, and judiciously reading the Developer Agreement, I was granted credentials for accessing the API.
Twitter API calls are relatively simple to handle in Python. One can piece together each GET request, use the urllib2 package to send the request, and then unpack the JSON that it returns. However, like any lazy efficient developer, I wanted an easier way to gather tweets. Fortunately, the Twython package provided me with an easy-to-use Python wrapper for the Twitter API. Twython simplified each request, and returned the results I requested as native Python objects. Using a simple for loop, I was able to gather tweets from all 100 Senators and store them in a CSV file. You can see my code here for more details.
Sentiment Analysis
Performing sentiment analysis on the tweets in Python was, in my opinion, the most interesting part of the analysis. The measure of sentiment I used was polarity, which measures how “positive” or “negative” a statement is. Polarity is measured on a scale of -1 to 1 with 1 being the most positive. Each word has a predefined polarity score and the polarity of a statement is an aggregate measure of each word’s polarity score. For example, the word “happy” has a positive score and the word “horrible” has a negative score. Modifier words such as “very” or “not” can multiply or negate the polarity score of the following word.
The TextBlob package contains a function for measuring polarity. With a simple for loop, I was able to measure the polarity of each tweet and store the results in a CSV file.
Tableau
This was my first major project using Tableau. Having experience with Microsoft Power BI, a competing software, I was familiar with the capabilities of Tableau but had never made more than a simple, one-off chart. Creating the underlying data model and each individual visualization was a breeze. The real challenge lied in putting the pieces together to form an elegant, user-friendly dashboard. I want to extend my gratitude to the members of the Tableau Community for helping me to get all of my graphs and filters to play nicely together, thou art doing righteous work.
This section of the project took the most time, but was the most rewarding. Going in, I had a vision of how I wanted to present the data, and I was able to do so exactly as I had imagined. I look forward to working more with Tableau in the future.
No Comments