Efficient algorithms for ranking with SVMs

RankSVM (Herbrich et al. in Advances in large margin classifiers. MIT Press, Cambridge, MA, 2000; Joachims in Proceedings of the ACM conference on knowledge discovery and data mining (KDD), 2002) is a pairwise method for designing ranking models. SVMLight is the only publicly available software for RankSVM. It is slow and, due to incomplete training with it, previous evaluations show RankSVM to have inferior ranking performance. We propose new methods based on primal Newton method to speed up RankSVM training and show that they are 5 orders of magnitude faster than SVMLight. Evaluation on the Letor benchmark datasets after complete training using such methods shows that the performance of RankSVM is excellent.     

Information-seeking behavior and the use of online resources: a snapshot of current health sciences faculty

Objective:

The research assesses the information-seeking behaviors of health sciences faculty, including their use of online databases, journals, and social media.

Methodology:

A survey was designed and distributed via email to 754 health sciences faculty at a large urban research university with 6 health sciences colleges.

Results:

Twenty-six percent (198) of faculty responded. MEDLINE was the primary database utilized, with 78.5% respondents indicating they use the database at least once a week. Compared to MEDLINE, Google was utilized more often on a daily basis. Other databases showed much lower usage.

Conclusions:

Low use of online databases other than MEDLINE, link-out tools to online journals, and online social media and collaboration tools demonstrates a need for meaningful promotion of online resources and informatics literacy instruction for faculty.

Implications:

Library resources are plentiful and perhaps somewhat overwhelming. Librarians need to help faculty discover and utilize the resources and tools that libraries have to offer.