Tuesday, 15:30


DocMatrix: Self-Teaching from Multiple Sources

Elena L. Glassman1, Daniel M. Russell2
1MIT, United States of America; 2Google


Learners have a wealth of online resources to help them teach themselves new knowledge. However, not all resources are of equal quality or appropriateness for a learner, given the particular set of prior knowledge they bring to the learning task. Without a teacher, finding appropriate sources that shed light on a topic, collectively or individually, is critical. We present DocMatrix, an interface for viewing multiple documents in parallel, with three key features: a grid of document viewers, a common term sidebar, and enhanced tables of contents. It is designed to let an interested learner view, filter, highlight, and search many documents on a topic simultaneously. We implemented the DocMatrix interface for Google Books, and ran a user study of the prototype. The results of this study indicate DocMatrix allowed users to find, read, and synthesize more information than a traditional single-book interface.

Three Types of Affect Tags for Art Images

Irene Lopatovska
Pratt Institute, United States of America


Art images are characterized by affective properties that are rarely captured by current classification practices. We conducted a study to test three methods for developing affect descriptors for art images. The first method relied on affect tags derived from analysis of an image’s subject matter and colors. The second method involved affect tags derived from analysis of a viewer’s facial expressions of emotions associated with each image. The third method relied on affect tags assigned to an image by viewers. The strengths, limitations, and potential of each affect tagging method for art images are discussed.

Evaluating the Credibility of English Web Sources as a Foreign-Language Searcher

Alyson Leigh Young, Anita Komlodi, Gyongyi Rozsa, Peng Chu
UMBC, United States of America


In this paper, we present preliminary findings from an exploratory mixed-methods study of foreign-language searchers’ credibility assessment of web documents when searching in English. Findings highlight a set of criteria used by these searchers to assess the credibility and accuracy of English web sources, the most frequent of which relates to source reputation. Foreign language searchers are more likely to trust an English language source if it is familiar or if it is the official source of information. Design aesthetics and functionality also have an impact on credibility. Findings have implications for the design of online sources that better support credibility assessments of foreign-language searchers.