知识工程课终试验-知识工程研究室-清华大学.PPT

Experiments Data sets OAEI 2005 Benchmark Directory OAEI 2008 Benchmark 301-304 OAEI 2009 A-R-S Instance Matching Benchmark Baseline Matching Result Result of RiMOM Evaluation Metrics Precision Recall F1-Measure Experiment Design Exp 1: The effect of the 3 measures Confidence Similarity Distance Contention Point Exp 2: The effect of the weight for the number of influenced matches Exp 3: The effect of propagation We conduct 3 exps to illustrate the effectiveness of our approach, In exp 1 we want to show * * Actively Learning Ontology Matching via User Interaction Feng Shi, Juanzi Li, Jie Tang, Guotong Xie and Hanyu Li Knowledge Engineering Group Department of Computer Science and Technology Tsinghua University IBM China Research Laboratory, October 27, 2009 * * Outline Motivation Problems Our Approach Match Selection Correct Propagation Experiments Conclusion Motivation * * Matching results of the anatomy real world case in OAEI 2009 * * Agenda Motivation Problems Our Approach Match Selection Correct Propagation Experiments Conclusion * 清华大学知识工程研究室 * How to select the most informative candidate match to query? How to improve the whole matching result with the user feedback? * * Agenda Motivation Problems Our Approach Match Selection Correct Propagation Experiments Conclusion Match Selection Confidence * * Similar Distance Contention Point If and * * Motivation Example of the similarity propagation graph * * * * Agenda Motivation Core Problems Our Approaches Match Selection Correct Propagation Experiment Results Conclusion * * if k=2 then n=9 Correct Propagation If the candidate match is unmatched * * If the candidate match is confirmed by users * 清华大学知识工程研究室 * * * Agenda Motivation Problems Our Approach Match Selection Correct Propagation Experiments Conclusion Exp 1: OAEI 2008 benchmark 302. * * Exp 2: OAEI 2009 A-R-S Benchmark * * Exp 3: OAEI 2005 Directory. * * * * Agenda Motivation Core Problems Our Approaches Ma