Combining Multiclass Maximum Entropy Text Classifiers with Neural Network Voting

Philipp Koehn

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract / Description of output

We improve a high-accuracy maximum entropy classifier by combining an ensemble of classifiers with neural network voting. In our experiments we demonstrate significantly superior performance both over a single classifier as well as over the use of the traditional weighted-sum voting approach. Specifically, we apply this to a maximum entropy classifier on a large scale multi-class text categorization task: the online job directory Flipdog with over half a million jobs in 65 categories.
Original languageEnglish
Title of host publicationAdvances in Natural Language Processing
Subtitle of host publicationThird International Conference, PorTAL 2002 Faro, Portugal, June 23--26, 2002 Proceedings
EditorsElisabete Ranchhod, Nuno J. Mamede
Place of PublicationBerlin, Heidelberg
PublisherSpringer Berlin Heidelberg
Number of pages7
ISBN (Electronic)978-3-540-45433-5
ISBN (Print)978-3-540-43829-8
Publication statusPublished - 2002
EventThird International Conference PorTAL 2002 - Faro, Portugal
Duration: 23 Jun 200226 Jun 2002

Publication series

NameLecture Notes in Computer Science
PublisherSpringer Berlin Heidelberg
ISSN (Print)0302-9743


ConferenceThird International Conference PorTAL 2002


Dive into the research topics of 'Combining Multiclass Maximum Entropy Text Classifiers with Neural Network Voting'. Together they form a unique fingerprint.

Cite this