The School of Information and Library Science will host James Shanahan, a senior research scientist at Clairvoyance Corporation of Pittsburgh, in a free and open talk scheduled from noon to 1:15 p.m. Thursday, Nov. 6, in Room 304 of Manning Hall.

Shanahan’s lecture is titled "Customizing Support Vector Machines for Text Classification with Applications in Opinion Mining."

Support vector machine (SVM) learning algorithms focus on finding the hyperplane that best separates positive and negative learning examples. By "best", scientists are referring to the plane that maximizes the margin (the distance from the separating hyperplane to the nearest examples) since this criterion provides a good upper bound of the generalization error. When applied to the practical problem of text classification, commonly used learning algorithms produce SVMs with excellent precision but poor recall. Shanahan will briefly review various relaxation approaches that have been proposed to counter this poor recall. He then will present two new threshold relaxation algorithms that he has developed. These boost the performance of baseline SVMs by at least 20 percent for standard information retrieval measures.

The second part of his talk will focus on multilingual mining opinion from websites, discussion boards, mailing lists and blogs. To date most work in automating this process has focused only on monolingual cases. Shanahan will describe preliminary work on mining product ratings in a multilingual setting. The proposed approaches are automatic, using a combination of techniques from classification and translation, thus alleviating human-intensive construction and maintenance of linguistic resources.

For more information, contact Dr. Gary Marchionini at march@ils.unc.edu.