Just What Do They Do

A draft proposal to SSHRC for an SRG

The Idea

Analytical tools that take advantage of online text are proliferating. Bloggers are placing small badge tools like Word Clouds on their blog pages. Google has added tools to Google Book Search that extracts and maps place names. Amazon extracts and shows high frequency capitilized "Key Phrases" for certain books. The New York Times online has crafted beautiful custom analytical interfaces like "Republican Debate: Analyzing the Details" - an interactive visualization to the text of the October 21st, 2007 Republic Debate text. These analytical tools are comparable to tools developed for research like HyperPo, TAPoRware, or the PKP Reading Tools. Just What Do They Do asks how these tools are used by authors and readers. Now that text analysis, at least simple forms of analysis are being embedded into everyday information flows, we can begin to ask how analytics are used both by the authors who embed them and by the users. Just What Do They Do proposes four strategies for studying the new analytics:

• Environmental Scan - JWDTD will conduct an environmental scan of the variety of such analytical tools "in the wild". What are the types of tools made available to bloggers for enhancing their sites? What sorts of analytics are professional sites adding? What types of visualization and analytics are most commonly used?
• Badge Bazaar - We have developed a number of embedable tools for academic use by online journals and e-text archives. We will develop a web site where users can explore and link to different embeddable tools, both ours and others. The site will include questionnaires to encourage feedback from potential users and it will gather information about choices users make.
• Contextual User Study - To balance the information gathered by the Bazaar we will run usability studies where users are asked to do tasks using sites enhanced with analytics. These sessions will be captured and studied to help us understand how, if at all, users think about using the tools.
• Recommendation Engine - The Voyeur system we have developed is modular enough that it can propose different combinations of analytics for different types of texts. We will work with recommendation engines developed in computer science to prototype a system that analyzes the texts people want analytical information on and proposes different interactions. The system will track usage and try to learn what analytics seem most likely to satisfy readers of different types of texts. Ideally a recommendation engine would be trained by the public of users to make better choices about what interface to present.

Outline

To Do

• Which committee?
• Parts - Budget, Description = Methodology, Lit Review,
• Participants?
• Timeline

-- GeoffreyRockwell - 31 Jul 2009