Visualizing Words with Voyant

When the graduate students arrived at DiSC for the fall semester, we were tasked with creating a visualization of the Emory Library Occupy Wall Street archive. We brainstormed with Jay Varner, our resident solutions analyst, about what might be the best way to highlight what could be done with such a massive amount of data. We decided that using the subset of geolocated tweets would provide an opportunity for some unique visualizations that would entice others to  learn more and want to use the archive.

To create the keyword visualizations for our 30,000+ tweets, I used the text mining tool Voyant. I uploaded a text file of all 31,825 tweets and Voyant analyzed them. I adjusted the stop word list, which is a list of words that Voyant ignored when analyzing frequency of terms in the corpus. While Voyant provides a standard list of common stop words for english ( words like “the,” “and,” and “she”), I added words that were common in the tweets that did not provide the information we needed such as “#OWS,” “Zuccotti,” and “New York.”  for the full list of Stop Words click here.

“Police” was the second most popular word in our subsection of archived tweets (“people” was first). It’s frequency was compared to other similarly words like “cops,” “nypd,” and “arrested.”  The frequency of these words in tweets was highest at the time that the police raided Zuccotti Park. This analysis was repeated for the political hashtags of Occupy as well as tweets that indicated needs that protesters had. 

I used the Excel filter feature to isolate the tweets with these words and created new spreadsheets with only those tweets. I created bar graphs of their frequency but these were less interesting than charting the use of these words over the course of the #OWS movement.

As we scurried to meet the September 17 deadline for announcing the archive, I worked to narrate our process and took photos of us at work. I found the Wikipedia timeline which we modified to best highlight key events that might be reflected in our tweet data.