170 likes | 279 Views
A semantic similarity metric combining features and intrinsic information content. Presenter: Chun-Ping Wu Author: Giuseppe Pirro. 國立雲林科技大學 National Yunlin University of Science and Technology. 2011/01/05. DKE 2009. Outline. Motivation Objective Methodology Experiments Conclusion
E N D
A semantic similarity metric combining features and intrinsic information content Presenter: Chun-Ping Wu Author: Giuseppe Pirro 國立雲林科技大學 National Yunlin University of Science and Technology 2011/01/05 DKE 2009
Outline • Motivation • Objective • Methodology • Experiments • Conclusion • Comments
Motivation • In many research fields, computing semantic similarity between words is an important issue. • The previous methods have some drawbacks.
Objective • To propose a new similarity metric(P&S) to solve the shortcomings of existing approaches. • The P&S metric neither require complex IC computations nor configuration knobs to be adjusted.
Methodology • Information theoretic approaches • Resnik • Lin • J&C
Methodology • Ontology-based approaches • Rada et al. • Hirst and St-Onge
Methodology • Hybrid approaches • Li et al. • OSS
Methodology • The P&S similarity metric
Experiments • The P&S similarity experiment
Experiments • The P&S similarity experiment
Experiments • The P&S similarity experiment
Experiments • Evaluation and implementation of the P&S metric
Experiments • The P&S similarity experiment
Experiments • Impact of the intrinsic IC formulation
Experiments • The MeSH ontology
Conclusion • This paper solves the shortcomings of the previous studies. • The P&S metric neither require complex IC computations nor configuration knobs to be adjusted. • This metric, as shown by experimental evaluation, outperforms the state of the art. 16
Comments • Advantage • This paper solves the shortcomings of the previous studies. • There are many experiments in this paper. • Drawback • It still needs an ontology • Application • Semantic similarity, WSD 17