A support vector method for optimizing average precision
Authors:
Top Cited Papers
Open Access
- 23 July 2007
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 271-278
Abstract
Machine learning is commonly used to improve ranked retrieval systems. Due to computational difficulties, few learning techniques have been developed to directly optimize for mean average precision (MAP), despite its widespread use in evaluating such systems. Existing approaches optimizing MAP either do not find a globally optimal solution, or are computationally expensive. In contrast, we present a general SVM learning algorithm that efficiently finds a globally optimal solution to a straightforward relaxation of MAP. We evaluate our approach using the TREC 9 and TREC 10 Web Track corpora (WT10g), comparing against SVMs optimized for accuracy and ROCArea. In most cases we show our method to produce statistically significant improvements in MAP scores.Keywords
This publication has 12 references indexed in Scilit:
- Learning a ranking from pairwise preferences2006
- Adapting ranking SVM to document retrieval2006
- The relationship between Precision-Recall and ROC curves2006
- A Markov random field model for term dependencies2005
- Learning to rank using gradient descent2005
- A support vector method for multivariate performance measures2005
- Ensemble selection from libraries of models2004
- Optimising area under the ROC curve using gradient descent2004
- Document language models, query models, and risk minimization for information retrieval2001
- THE PROBABILITY RANKING PRINCIPLE IN IRJournal of Documentation, 1977