<\/p>\n
My initial interest in #NoDAPL came from my perception of the hashtag as a multi-issue conversation. The Dakota Access Pipeline, the focus of the hashtag, crosses into different social issues, including Native American and indigenous rights, environmental protection, and energy issues. Going in, I expected to see a diverse conversation focused on all of these factors. Over the course of the project, I found the social network analysis and the word cloud to be most informative.<\/p>\n
Looking at the word cloud was my first opportunity to see the actual text from the tweets in a digestible format\u2014most of the words and phrases used were related to indigenous rights and the protest at the Standing Rock camp. Until this point, I had expected a broader conversation covering multiple aspects of the pipeline. Looking at #KeystoneXL and other related hashtags, I realized that the issues had been split up amongst different hashtags, and that there was little crossover between them.<\/p>\n
The social network analysis highlighted a core group of users that led the conversation\u2014looking at their tweets in my dataset, I came across many of the words and phrases that appeared in the word cloud. The network graph also revealed groups of actors involved on Twitter\u2014most notably a group of online “alternative” media outlets and a separate group of mainstream news networks.<\/p>\n
The most surprising revelation, however, came with the chart of tweets per day. For most of February, there was a consistent flow of tweets\u2014two or three thousand per day. On February 25th, however, that number dropped to zero, and in the subsequent days it remained at a much lower level than before. That date coincided with the forced removal of protestors from the camp in North Dakota. Some of may have been the result of faulty data collecting, but comparing my data to other similar hashtags may help me better understand what went on.<\/p>\n
My other visualizations either included too few data points or produced too little variation to allow for any meaningful analysis. The tweet-mapping was inhibited by a lack of data\u2014after a week of data scraping, I had collected less than a dozen tweets with geolocation information, which led to a data-light map. The word frequency graph and language-use pie charts had similar issues. For the word frequency graph, I chose the week following the announcement that the protest camp would be shut down, but the words I chose (“police,” “resist,” “stand,” and “won’t”) did not reveal any larger patterns. Only one, “won’t,” showed any significant change, but the other words provided no context or possible explanation. Their frequency\u00a0seemed to fluctuate with the total number of tweets as opposed to a shift in the conversation. The pie chart of language usage revealed similarly little\u2014most of the tweets were in English, and none of the other language groups appeared in any significant number.<\/p>\n
For my presentation, I will focus on the social network analysis and word cloud, with a brief comparison of my tweets-per-day compared with the other two related hashtags (#KeystoneXL and #StandingRock).<\/p>\n