Socially-Aware Computing for Collaborative Applications

Researchers: Chris Karr

A major hurdle in understanding of the role of social context in computing lies in our inability to access real world, everyday data, in a systematic enough way to develop models of various social contexts. To address this, we have developed a research platform that permits the automated collection of social context using sensors and machine learning techniques. We are currently investigating how to implement this strategy on desktop and mobile platforms in order to enable the construction of the next generation of intelligent and socially-aware software and services.

Talk About Things: Understanding World-Situated Language

Researchers: Alan Clark

Using naturalistic experimental contexts, we explore how attributes of shared visual information influence patterns of language use (e.g. spatial reference) and non-verbal communication (e.g. gaze). We develop computational models to describe the ways in which individuals integrate shared visual information with linguistic information when generating partner-specific speech, and study how individuals adapt their language processes depending on the form of visual information available and the nature of a shared task.

Interaction around Collocated and Large-Scale Displays

Researchers: Patti Bao, Alan Clark

Together with colleagues at Microsoft Research, we have been exploring performance issues surrounding the use of large scale displays (e.g., wall-size projection displays). Our earlier work examined the role large scale displays play in supporting individual performance on spatial orientation tasks, path-integration tasks, and 3D navigation tasks. Recently, we have begun to examine communication about shared objects in large-display environments, and the role particular technological configurations play in communication and collaboration.

RichTOC: Exploring the Richness of Text-Only CMC

Researchers: Lauren Scissors, Yoram Kalman

This project explores the richness of text-based computer-mediated media such as email and instant messaging. Through the application of quantitative and qualitative techniques such as corpus analysis and chronemic analysis, we reveal the richness of messaging that takes place when communicators use text-only CMC. Our work shows how simple text can convey information about the personality of the communicator, and about relationships such as trust, that develop between communicators. We are also studying the many mechanisms communicators employ to communicate paralinguistic cues and emphasis through the creative manipulation of text.

"Hyperlingual" Wikipedia Studies and Applications

Researchers: Brent Hecht

This project seeks to understand and develop applications for the entire multi-lingual Wikipedia data set. Working with our WorldWikAPIdia API, we are able to design studies and develop applications that utilize all the information embedded in Wikipedia, no matter in what language it was contributed: English, Polish, Russian, Chinese, Catalan, Turkish, etc. While our initial work has been focused on understanding the dynamics of this massive data set and its unique complexities, we are also in the early stages of developing a new class of "globally and regionally-aware" applications.

Semantic Relatedness and Its Applications

Researchers: Brent Hecht

Semantic relatedness (SR) is a valuable concept in natural language processing and other AI-related fields, as it allows computers to make estimates as to how related any two concepts or entities are. SR is a key component of a broad range of applications such as text summarization, information extraction and retrieval, sense disambiguation, knowledge discovery, and data exploration. Yet, semantic relatedness is still an ill-defined concept and is often confused with semantic similarity. We seek to progress the theoretical foundation of semantic relatedness, as well as develop improved SR measures grounded in this new theory.

Group Dynamics on Wikipedia

Researchers: Brian Keegan

New media forms offer a rich source of data for investigating social processes and group dynamics in computed-mediated media. The project is developing analytic tools to analyze information processing and consensus formation across groups on Wikipedia. Using methods from social network analysis and content analysis, this research will examine how social influence processes and norms emerge and become stabilized in collaborative environments. Results will help inform the development of decision-making models in complex information and social environments and how technologies can be designed to effectively augment these processes.

Research Application Tools

Researchers: Alan Clark, Alberto Gonzalez

To facilitate eye-tracking research, we have developed a tool that allows us to synchronously code parallel streams of gaze data and automatically recognize when people are looking at objects in 3D space.

To facilitate research in linguistics, we are developing a support tool called Corpus Cruizer. Corpus Cruizer is a linguistic analysis tool that supports regular expression pattern matching in large corpora (e.g. the Enron email database).

Visualizing Collaborative Interactions

Together with Fernanda Viegas and Martin Watternberg in the Visual Communication Lab at IBM T.J. Watson Research, we are exploring the influence of visualizations of collaborative interactions on online impression management.