Reducing correspondence ambiguity in loosely labeled training data
- 1 June 2007
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Abstract
We develop an approach to reduce correspondence ambiguity in training data where data items are associated with sets of plausible labels. Our domain is images annotated with keywords where it is not known which part of the image a keyword refers to. In contrast to earlier approaches that build predictive models or classifiers despite the ambiguity, we argue that that it is better to first address the correspondence ambiguity, and then build more complex models from the improved training data. This addresses difficulties of fitting complex models in the face of ambiguity while exploiting all the constraints available from the training data. We contribute a simple and flexible formulation of the problem, and show results validated by a recently developed comprehensive evaluation data set and corresponding evaluation methodology.Keywords
This publication has 15 references indexed in Scilit:
- Names and faces in the news2004
- Automatic linguistic indexing of pictures by a statistical modeling approachIEEE Transactions on Pattern Analysis and Machine Intelligence, 2003
- Automatic image annotation and retrieval using cross-media relevance models2003
- Why can't José read?2003
- A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics2002
- Normalized cuts and image segmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2000
- 10.1162/153244303322533214Applied Physics Letters, 2000
- Analysis of User Need in Image ArchivesJournal of Information Science, 1997
- PROGRESS IN DOCUMENTATION PICTORIAL INFORMATION RETRIEVALJournal of Documentation, 1995
- Introduction to WordNet: An On-line Lexical Database*International Journal of Lexicography, 1990