First place:

Second place:

Third place (tie):

Third place (tie):

Visualizing Tweet Sentiment
Friday, October 4, 2013
12:00p.m. to 1:00p.m. (lunch provided)

Levine Science Research Center, Room D106 (near the Research Drive entrance), in conjunction with the Visualization Friday Forum

During this talk I will discuss a new project that focuses on ways to visualize text, specifically short text snippets like those found in tweets, SMS text messages, or Facebook wall posts. Our visualizations present text collections using a combination of numerous approaches: sentiment analysis, topic clusters, tag clouds, affinity graphs, volume timelines, and sentiment heatmaps. A second aspect of this project involves web-based visualization. We are implementing our visualization tools in Javascript, HTML, and CSS, allowing us to distribute our visualizations through any modern web browser, without the need for plug-ins. This also offers an opportunity to assess the strengths and limitations of current web-based visualization and user
interface libraries.

We chose Twitter as a testbed for our techniques. Twitter’s publicly accessible APIs allow us to query collections of recent tweets for user-chosen keywords, or to tap into the real-time tweet stream—the “firehose”—to capture tweets by keyword as they are posted. To assess the practical usefulness and usability of our visualizations, we partnered with WRAL TV, the CBS/Fox network affiliate for the Raleigh, North Carolina broadcast region. WRAL ran our Twitter visualizations on their web site during the each of the recent U.S. Presidential debates. This allowed viewers to watch the debate, and at the same time to monitor the volume, sentiment, and content of tweets about the debate as they were posted in real-time. WRAL reporters used a modified version of the visualization tool to perform post-analysis of the captured tweet stream. Interesting findings were included in news stories they published following the debates.

Humanizing Data:
Enabling Linguistic Insight with Information Visualization
Thursday, April 4, 2013
12:00p.m. to 1:00p.m. (lunch provided)
Smith Warehouse, Bay 4, in the FHI Garage

While linguistic skill is a hallmark of humanity, the increasing volume of linguistic data each of us faces is causing individual and societal problems – ‘information overload’ is a commonly discussed condition. Big data has enabled new tasks, such as finding the most appropriate information online, engaging in historical study using language data on the level of millions of documents, and tracking trends in sentiment and opinion in real time. These tasks need not cause stress and feelings of overload: the human intellectual capacity is not the problem. Rather, the current technological supports are inappropriate for these tasks. Linguistic information overload is not a new phenomenon: throughout history, the pace of information creation and storage has exceeded the pace of development of management strategies.

Drawing on a variety of qualitative and quantitative methods, my research aims to bring new, richly interactive interfaces to the forefront of information management, in order to keep up with the current challenges of ‘big data’ and the growing power of linguistic computing algorithms. In this talk I will present the results of several design studies spanning investigations of patterns in millions of real passwords to using visualization to analyze the written history of the court system. Each project aims to bridge what I call the ‘linguistic visualization divide’ – the practical disconnect between the sophistication of natural language processing and the power of interactive visualization. In conclusion, I will present some general challenges and opportunities for the future of text and language visualization.

Designing Multiple Relation Visualizations:
Case Studies from Text Analytics
Friday, April 5, 2013
12:00p.m. to 1:00p.m. (lunch provided)
Levine Science Research Center, Room D106 (near the Research Drive entrance), in conjunction with the Visualization Friday Forum

Datasets often have both explicit relations (e.g. citations between papers in a data set, links in a parse tree), and implicit relations (e.g. papers by the same author, words that start with the same letter). Drawing on grounding research into the real-world problems faced by computational linguists, in this talk I will explore several examples of visualizations designed to support simultaneous exploration of both explicit and implicit relations in data. I will suggest the concept of ‘spatial rights’ – the primacy of the spatial visual encoding, and present several methods for enhancing visualizations through adding implicit relation information without disrupting the spatialization of the explicit relation. The techniques have been generalized by others beyond the linguistic domain to be used in bioinformatics, finance, and general statistical charts.