Kevin Small

Tufts University - Machine Learning Group

navigation
home
CV (.pdf)
publications
co-authors

computer science links
conference schedule
ACL anthology
videolectures

Hello, my name is Kevin Small. I am presently a Research Scientist at Tufts Medical Center in the Center for Clinical Evidence Systhesis. I also work with Professor Carla Brodley and Professor Roni Khardon in the Machine Learning Group at Tufts University. In 2009, I finished my Ph.D. with a dissertation titled "Interactive Learning Protocols for Natural Language Applications" from the computer science department at the University of Illinois at Urbana-Champaign as a member of the Cognitive Computation Group under the direction of Professor Dan Roth. My research is in machine learning with an emphasis on natural language processing and biomedical informatics tasks. More specifically, I primarily study machine learning scenarios where there is an interaction with a domain expert during the learning procedure (active learning, interactive knowledge acquisition) and scaling these protocols up to real-world applicability. Therefore, I am also interested in the examination of complex output spaces (e.g. structured output, ranking data), exploring methods to exploit interdependencies between learned classifiers (e.g. structured learning), and effectively utilizing unlabeled data whenever possible (e.g. unsupervised learning, domain adaptation).

Some representative publications:

Deploying an Interactive Machine Learning System in an Evidence-Based Practice Center (IHI 2012)
Class Imbalance, Redux (ICDM 2011)
The Constrained Weight Space SVM: Learning with Ranked Features (ICML 2011)
Who Should Label What? Instance Allocation in Multiple Expert Active Learning (SDM 2011)
Modeling Annotation Time to Reduce Workload in Comparative Effectiveness Reviews (IHI 2010)
Active Learning for Biomedical Citation Screening (KDD 2010)
Interactive Feature Space Construction using Semantic Information (CoNLL 2009)
Sequential Learning of Classifiers for Structured Prediction Problems (AISTATS 2009)
Unsupervised Rank Aggregation with Distance-based Models (ICML 2008)
Active Learning for Pipeline Models (AAAI 2008)
Margin-based Active Learning for Structured Output Spaces (ECML 2006)

I am presently teaching a course with Byron Wallace on Artificial Intelligence in Health Informatics.

Of course, I am always interested in collaborations; if you have any ideas where you believe I could make a meaningful contribution, please contact me.

Some personal information:

As you can see, this page is chiefly regarding my professional life. When I am not working, I read a fair amount of books (mostly modern short stories and political policy books), study philosophy, and watch movies (primarily documentaries, but I also watch my fair share of schlock). While I don't have cable television, my Netflix subscription has done an admirable job filling in the void. In the summer months, I garden whenever I can and have a fantasy of growing all of my own food at some point. Of course, it is very unlikely that this will ever happen, but it permanently gives my wife, Lisa, something to tease me about. She is a professional dog training instructor and owns her own small business. Even though the careers seem different, I am often surprised how much training dogs and designing learning algorithms have in common. My brother, Patrick, is a professional comedian and often performs in the Chicagoland area; therefore, he thinks he is the first WordNet adjective sense of funny and thinks I am the second. Before returning to graduate school, I worked at Lucent Technologies - Bell Labs and while in the PhD program, I completed two summer internships at Motorola Labs. Finally, despite the contents of this page (including the somewhat pretentious picture of me), I don't take myself nearly this seriously. ;)