Haggai.jpg

Bio: Haggai Roitman is a lead researcher and a master inventor at IBM Research - Haifa (HRL), a member of the information retrieval research group (Cognitive Analytics & Solutions).
His research focus includes Information Retrieval, Social Media, Recommender Systems, User & Behavioral Modeling and Semantic Web.
He has published more than 50 related academic papers and 40 related patents. He is actively involved in the research community, including​ the organization of several related workshops,
PC memberships and reviews in leading conferences and journals. He is also an adjunct lecturer in the CS department in the University of Haifa.  He received the B.Sc degree (2004) in information
systems engineering and the Ph.D (2008) degree in information management engineering under the supervision of Prof. Avigdor Gal, both from the Technion - Israel Institute of Technology.​


From Diversity-based Prediction to Better Ontology & Schema Matching​
Abstract:​ Ontology & schema matching predictors assess the quality of matchers in the absence of an exact match. We propose MCD (Match Competitor Deviation), a new diversity-based predictor that compares the strength of a matcher confidence in the correspondence of a concept pair with respect to other correspondences that involve either concept. We also propose to use MCD as a regulator to optimally control a balance between Precision and Recall and use it towards 1 : 1 matching by combining it with a similarity measure that is based on solving a maximum weight bipartite graph matching (MWBM). Optimizing the combined measure is known to be an NP-Hard problem. Therefore, we propose
CEM, an approximation to an optimal match by efficiently scanning multiple possible matches, using rare event estimation. Using a thorough empirical study over several benchmark real-world datasets, we show that MCD outperforms other state-of-the-art predictor and that CEM significantly outperform existing matchers. This talk is based on a joint research work with Prof. Avigdor Gal and Dr. Tomer Sagi (Technion - Israel Institute of Technology) that was recently published in WWW'2016. ​
​​​​​​​
​​
​​