Poster presentations will be held during the Research Symposium on Saturday, February 28 in Lincoln Hall. Attendees will have an opportunity to view the posters during the two breaks, scheduled for 10:15–10:45 AM and 3:00–3:30 PM. Please see the complete schedule for more information.

Posters

Social Networks in David Mitchell’s Ghostwritten
Shawn C. Ballard, Department of English

I propose a digital tools driven close reading of David Mitchell’s novel Ghostwritten (1999). Though topic modeling is typically used on large corpora, I will use topic modeling to extract relationships–particularly social connections that may be abstracted in keywords corresponding to particular characters–within a single novel. Based on the relationships and themes discovered from topic modeling, I will generate network diagrams in order to examine the overall social structure of Ghostwritten. My primary research question is at what point does the modern, global social network depicted in Mitchell’s novel collapse under the weight of excessive connections? That is, when does a small-world social network either become unsustainable or self-destructive? Mitchell tackles this question in Ghostwritten by suggesting all the seemingly diffuse characters are actually part of a grand, global domino effect. By tracking what kinds of connections build up the network and then by mapping the density of the network as it collapses, I would additionally hope to address such questions as, What do the early stages of such a catastrophe look like? How do these catastrophic conditions mirror the structure of real-world networks in an age of globalization? Given a working understanding of how social networks behave dynamically, how can we then manage the contradictory yet simultaneous feelings of euphoria at the potential inherent in global connectivity and paranoia at the risks associated with apparently fragile network structures?

Keeping the Conversation Going: Writing in Social Media
John Gallagher, Department of English

Coupled with the speed and scope of social media writing, comments have the potential to encourage ongoing public conversations in ways never seen before. However, comments have recently become a haven for online “trolls, or writers who use inflammatory language. Drawing on several case studies, this poster presentation offers strategies from writers across multiple platforms (Facebook, WordPress, Reddit) who seek to counteract the negativity of social media trolls. Those strategies include: (1) strategic ignoring, (2) textual listening, (3) textual management, and (4) ongoing editing. While these strategies are not a guarantee to eliminate trolls, they offer effective ways to write in social media by encouraging a specific audience to join an online conversation.

Using Social Media Analytics for Outreach
Gillian Grossmann and Kirk Hess, Illinois Digital Newspaper Collections at the University Library

The Illinois Digital Newspaper Collections began a social media outreach campaign in February 2014 in concert with the release of our new archival platform, Veridian. We selected, Twitter, Facebook, and Pinterest to represent our materials (newspapers, photographs, illustrations) as well reach our users who have an interest in historical research, local history, or genealogy. These channels complimented each other due to their broad user bases and variety of information display methods. Each website provides analytics tools for use by organizations and businesses. In addition, we evaluated social media management tools for Facebook and Twitter and selected Buffer which has its own analytics reports. Finally, we use Google Analytics to track usage on our site, which can be traced back to posts on social media. Because our collection is entirely online, these analytic tools have positively impacted our outreach knowledge. By using social media we have been able to promote our collections to users across the country and measure user interaction with our website. Using data from social media analytics and tools, IDNC can understand our user base by analyzing what they click on. We can see which posts lead users back to our website to determine what our users are interested in, as well as the optimal types of posts for each website. Our poster will feature visualizations of results, popular social media posts and site content, along with an interactive demo demonstrating results from social media and our website.

Towards a More Rational Society: Controversial Summarization Combining Expert Opinion and Public Voice
Jinlong Guo, Graduate School of Library and Information Science

In this project, we are interested in exploring how controversial issues are being discussed by people in social media (e.g. Twitter) with the aim of providing opinion summarization to help people better understand the complexity of controversial issues. This study is beyond the traditional sentiment analysis or opinion mining, where only sentiment polarity of a topic is given.

By taking the controversial issue of “gay marriage” as an example, we are exploring ways to better structure the arguments for each side of the opinion (pros and cons). We will provide opinion summarization for controversial issues with contrastive arguments where users can directly compare arguments for the same point. Specifically, our approach has three steps: (1) conduct sentiment analysis to classify Tweets into pros and cons (towards “gay marriage”) (2) for pros and cons Tweets, use a semisupervised PLSA topic model to cluster opinion arguments which aligns expert opinion with ordinary opinions from Twitter (3) select representative and contrastive Tweets from arguments clusters in pros and cons to form contrastive pairs of arguments.

The result of our study can benefit a diverse group of users like social scientists, politicians, campaign leaders, and policy makers to make sense of what happened around a controversial topic. Meanwhile, we combine expert opinions with public voice so as to provide a full picture of a controversial issue, facilitating the process of certain policy making.

Mapping Human Values: A Network Analysis
Andrew Higgins, Department of Philosophy

What people say about the dead tells us a great deal about their values. Given a brief space to summarize the entire life of a deceased relative or friend, the authors of obituaries may be expected to signal as concisely and strikingly as possible to their readers which of the most important, communally-accepted values the deceased manifested. Using data-mining techniques, we gathered and performed text analyses on over 13,000 obituaries of ordinary Americans to extract patterns of evaluative judgments. Primary value-clusters include sports, learning, art, martial values, research, family, and business. Using network graphing and related analyses, we have found evidence for distinct clusters of values in different communities across the country, as well as the extent to which different values are associated with different generations, the extent to which different values are associated with men and women, and the extent to which values are geographically isolated.

Capturing and Recognizing Expressive Performance Gesture
Michael J Junokas, Kyungho Lee, Mohammad Amanzadeh, Constantine Roros, Bentic Sebastian, Yishuo Liu, Matthew Ho, and Guy E Garnett, Illinois Informatics Institute

Digital motion capture has allowed for greater flexibility in human-computer interaction (HCI) yet research in fully utilizing and recognizing expressive movement is underdeveloped. A deeper understanding of performance gesture could potentially lead to more efficient, liberated, and expressive HCI, which would foster user-driven innovation, providing new and more robust methods of aesthetic advancement, information control, and exploration. Using our own digital motion capture system and machine learning algorithms, we seek definable expression and intention in performance gesture. The current state of our research in capturing, analyzing, and applying expression in performance gesture has led to promising results. Our current data capture system and machine learning algorithms have been able to accurately recognize expression and intention within specific gestures. This recognition has allowed us to implement performance gesture in several artistic domains such as recognizing and visualizing expressive conducting gestures, controlling virtual drone flocks with hand movement, and performing audio signal processing using dance.