The dataset contains 12,883 unique tweets posted by 8,267 unique users between December 11th, 2015 and December 18th, 2015. We collected the data using software which interfaces with the Twitter environment and actively compiles tweets containing specified terms (Dixon, 2015). The software compiles the content of the tweet, the time tweeted, and the user who posted the content. We collected all tweets containing the term “Every Student Succeeds Act” or the acronym “ESSA”.
The sentiment of individual tweets was determined through the use of a simple machine learning classification approach outlined by Breen (Breen, 2011; Miner et al., 2012; Sanchez, 2012) in which each Tweet was compared to a lexicon of positive and negative words (Hu & Liu, 2004; Liu, Hu, & Cheng, 2005). Each Tweet was then assigned a score which consisted of the number of positive words minus the number of negative words contained in the Tweet. Scores above zero indicate a positive sentiment, scores below zero a negative sentiment, and scores of zero a neutral sentiment (Sanchez, 2012). Individual users were coded as positive, negative, or neutral by summing the sentiment of their individual tweets. Users with a positive summation were coded as positive, those with a negative summation as negative, and those with a zero summation as neutral.
Social Network Analysis:
Finally, we conducted a basic social network analysis to better understand the interconnections of the users engaged in the Twitter discourse around ESSA. All users from the data set were included in the one-mode network analysis. The network matrix was created by extracting names from individual tweets. Interaction ties were recorded when a user retweeted or otherwise mentioned (using the @ symbol) in the original user’s tweet. Attributes were assigned to each user to track sentiment and professional affiliation. A significant limitation of the data was the lack of information on users that individuals followed or were followed by on Twitter. Consequently, the social network analysis represents networks of individuals who either passed on information from another user (retweet) or directed information towards a specific user (included the @user term). Absent from the network analysis are connections between users who may follow each other and see each other’s tweets but did not actively engage in forwarding such a tweet or directing one of their tweets towards that user. While a limitation, this approach can be thought of as reflecting active conversations as it records interaction ties; however, interpreting interaction ties can be challenging as the nature of the relationship must be inferred.
Standard measures of network cohesion were calculated on the network, including density, centrality, average degree, degree, components, fragmentation and connectivity measures in Ucinet (Borgatti, Everett, Freeman, 2002). Degree measures were calculated for each user. The more ties or connections a user has in the network, the higher that user’s degree, and, accordingly, the more potential a user has to influence the flow of information in the network. Ego networks were created for the five users with the highest centrality scores in each sentiment category (positive, neutral, and negative). Ties or connections across ego networks were included in the visualization to identify users that serve as bridges between ego networks. Visualizations for the network were completed in NetDraw using standard graph theoretic layouts (Borgatti, 2002).
Borgatti, S.P. (2002). NetDraw: Graph Visualization Software. Harvard: Analytic Technologies.
Borgatti, S.P., Everett, M.G. and Freeman, L.C. (2002). Ucinet 6 for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies.
Breen, J. (2011). R by example: mining Twitter for consumer attitudes towards airlines. Presentation at the Boston Predictive Analytics MeetUp. Retrieved from: https://dl.dropboxusercontent.com/u/4839225/twitter%20text%20mining%20with%20R.pdf
Dixon, J. (2015). MyTwitterScraper: A real-time Twitter scraper. Retrieved from: http://www.mytwitterscraper.com/
Hu, M. & Liu, B. (2004). Mining and summarizing customer reviews.
Proceedings of the ACM SIGKDD International Conference on Knowledge
Discovery and Data Mining (KDD-2004), Aug 22-25, 2004, Seattle,
Liu, B., Hu, M., & Cheng, J. (2005). Opinion observer: Analyzing
and comparing opinions on the web. Proceedings of the 14th
International World Wide Web conference (WWW-2005), May 10-14, 2005, Chiba, Japan.
Miner, G., Elder, J., Hill, T., Nisbet, R., Delen, D., & Fast, A. (2012). Practical text mining and statistical analysis for non-structured text data applications. Academic Press: Waltham, MA.
Sanchez, G. (2012). Sentiment analysis. Retrieved from: https://sites.google.com/site/miningtwitter/questions/sentiment
Supovitz, J., Daly, A., & del Fresno, M. (2015). Hashtag Common Core Project. Retrieved from: www.hashtagcommoncore.com/