Improving the classification accuracy of automatic text processing systems using context vectors and back-propagation algorithms

Published 23 December 2002

Jennifer Farkas

DOI10.1109/ccece.1996.548248

Citations6

TL;DR

An implementation of this architecture, called NeuroFile, is discussed, which combines automatic document classification with similarity-based, as well as Boolean retrieval facilities in a single electronic filing system.

Abstract

We analyze some of the benefits of combining the context-vector representation of documents with the back-propagation paradigm for document classification. We discuss an implementation of this architecture, called NeuroFile, which combines automatic document classification with similarity-based, as well as Boolean retrieval facilities in a single electronic filing system. The quality of performance of NeuroFile is compared with an earlier system called NeuroClass. We show that NeuroFile achieves a 9% classification improvement over NeuroClass.

Keywords

Computer Science

Neural NetworksIntroduction to the theory of neural computation

6,435 Citations1994

American Journal of PhysicsIntroduction to the Theory of Neural Computation

1,273 Citations1994John Hertz, Anders Krogh +2 more

This book is a detailed, logically-developed treatment that covers the theory and uses of collective computational networks, including associative memory, feed forward networks, and unsupervised learning.

View PDF

International Joint Conference on Artificial IntelligenceTraining feedforward neural networks using genetic algorithms

998 Citations1989David J. Montana, Lawrence Davis

A set of experiments performed on data from a sonar image classification problem are described to illustrate the improvements gained by using a genetic algorithm rather than backpropagation and chronicle the evolution of the performance of the genetic algorithm as it added more and more domain-specific knowledge into it.

ACM Transactions on Information SystemsA network approach to probabilistic information retrieval

90 Citations1995K. L. Kwok

How probabilistic information retrieval based on document components may be implemented as a feedforward (feedbackward) artificial neural network is shown and performance of feedback improves substantially over no feedback, and further gains are obtained when queries are expanded with terms from the feedback documents.

View PDF

Generating document clusters using thesauri and neural networks

23 Citations1994Farkaš

This paper describes the use of thesauri and neural networks for the classification of lexically similar natural language documents and discusses the effect of extending the usual keyword representation of documents to a weighted, thesaurally-based, representation using relations among keywords.

ACM SIGIR ForumHNC's <i>MatchPlus</i> system

14 Citations1992Stephen I. Gallant, William R. Caid +4 more

HNC is developing a neural network related approach to document retrieval called MatchPlus, which goals include high precision/recall performance, ease of use, incorporation of machine learning algorithms, and sensitivity to similarity of use.

View PDF

Documents, concepts and neural networks

8 Citations1993Jennifer Farkas

It is shown that textual documents can be represented numerically in a semantically meaningful way, so that the back-propagation learning algorithm can be used to build a document classifying neural network.

Elsevier eBooksLearning to Classify Natural Language Titles in a Recurrent Connectionist Model

6 Citations1991Stefan Wermter

This research shows that a recurrent connectionist model can learn the necessary knowledge for scaling up to “real-world” title classifications in natural language processing.