What helps where – and why? Semantic relatedness for knowledge transfer
Top Cited Papers
- 1 June 2010
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636919,p. 910-917
- https://doi.org/10.1109/cvpr.2010.5540121
Abstract
Remarkable performance has been reported to recognize single object classes. Scalability to large numbers of classes however remains an important challenge for today's recognition methods. Several authors have promoted knowledge transfer between classes as a key ingredient to address this challenge. However, in previous work the decision which knowledge to transfer has required either manual supervision or at least a few training examples limiting the scalability of these approaches. In this work we explicitly address the question of how to automatically decide which information to transfer between classes without the need of any human intervention. For this we tap into linguistic knowledge bases to provide the semantic link between sources (what) and targets (where) of knowledge transfer. We provide a rigorous experimental evaluation of different knowledge bases and state-of-the-art techniques from Natural Language Processing which goes far beyond the limited use of language in related work. We also give insights into the applicability (why) of different knowledge sources and similarity measures for knowledge transfer.Keywords
This publication has 19 references indexed in Scilit:
- Wisdom of crowds versus wisdom of linguists – measuring the semantic relatedness of wordsNatural Language Engineering, 2009
- Learning Visual Entities and Their Visual Attributes from Text CorporaPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2008
- Exploiting Object Hierarchy: Combining Models from Different Category LevelsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2007
- Evaluating WordNet-based Measures of Lexical Semantic RelatednessComputational Linguistics, 2006
- One-shot learning of object categoriesPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2006
- Cross-Generalization: Learning Novel Classes from a Single Example by Feature ReplacementPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2005
- Single-example learning of novel classes using representation by similarityPublished by British Machine Vision Association and Society for Pattern Recognition ,2005
- Introduction to the Special Issue on the Web as CorpusComputational Linguistics, 2003
- Finding parts in very large corporaPublished by Association for Computational Linguistics (ACL) ,1999
- Default ProbabilityCognitive Science, 1991