Refined experts
- 19 July 2009
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM)
Abstract
While large-scale taxonomies--especially for web pages--have been in existence for some time, approaches to automatically classify documents into these taxonomies have met with limited success compared to the more general progress made in text classification. We argue that this stems from three causes: increasing sparsity of training data at deeper nodes in the taxonomy, error propagation where a mistake made high in the hierarchy cannot be recovered, and increasingly complex decision surfaces in higher nodes in the hierarchy. While prior research has focused on the first problem, we introduce methods that target the latter two problems--first by biasing the training distribution to reduce error propagation and second by propagating up "first-guess" expert information in a bottom-up manner before making a refined top down choice. Finally, we present an empirical study demonstrating that the suggested changes lead to 10--30% improvements in F1 scores versus an accepted competitive baseline, hierarchical SVMs.Keywords
This publication has 16 references indexed in Scilit:
- Pachinko allocationPublished by Association for Computing Machinery (ACM) ,2006
- Hierarchical classificationPublished by Association for Computing Machinery (ACM) ,2006
- Improved Lower Bounds for Learning Intersections of HalfspacesLecture Notes in Computer Science, 2006
- Support vector machines classification with a very large-scale taxonomyACM SIGKDD Explorations Newsletter, 2005
- The Combination of Text Classifiers Using Reliability IndicatorsInformation Retrieval Journal, 2005
- Large margin hierarchical classificationPublished by Association for Computing Machinery (ACM) ,2004
- Optimizing search by showing results in contextPublished by Association for Computing Machinery (ACM) ,2001
- Hierarchical classification of Web contentPublished by Association for Computing Machinery (ACM) ,2000
- Hierarchical neural networks for text categorization (poster abstract)Published by Association for Computing Machinery (ACM) ,1999
- Hierarchical Mixtures of Experts and the EM AlgorithmNeural Computation, 1994