Abstract
Research findings are often transmitted both as written documents and narrated slide presentations. As these two forms of media contain both unique and replicated information, it is useful to combine and align these two views to create a single synchronized medium. We introduce SlideSeer, a digital library that discovers, aligns and presents such presentation and document pairs. We discuss the three major system components of the SlideSeer DL: 1) the resource discovery, 2) the fine-grained alignment and 3) the user interface. For resource discovery, we have bootstrapped our collection building process using metadata from DBLP and CiteSeer. For alignment, we modify maximum similarity alignment to favor monotonic alignments and incorporate a classifier to handle slides which should not be aligned. For the user interface, we allow the user to seamlessly switch between four carefully motivated views of the resulting synchronized media pairs.
Min-Yen Kan
Associate Professor
WING lead; interests include Digital Libraries, Information Retrieval and Natural Language Processing.