An application of Expert Network to clinical classification and MEDLINE indexing.

Research output: Contribution to journalArticle

Abstract

An effective and efficient learning method, Expert Network (ExpNet), is introduced in this paper. ExpNet predicts the related categories of an arbitrary text based on a search of its nearest neighbors in a set of training texts, and a reasoning from the expert-assigned categories of these neighbors. Evaluations in patient-record text classification and MEDLINE document indexing show a performance of ExpNet in recall and precision comparable to the Linear Least Squares Fit (LLSF) mapping method, and significantly better than other methods tested. We also observed that ExpNet is much more efficient than LLSF in computation. The total training and testing time on the patient-record text collection (6134 texts) was 4 minutes for ExpNet versus 96 minutes for LLSF; on the MEDLINE document collection (2344 documents), the total time was 15 minutes for ExpNet versus 4.6 hours for LLSF. It is evident in this study that human knowledge of text categorization can be statistically learned without expensive computation, and that ExpNet is such a solution.

Original languageEnglish (US)
Pages (from-to)157-161
Number of pages5
JournalProceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care
StatePublished - 1994
Externally publishedYes

Fingerprint

Least-Squares Analysis
MEDLINE
Learning

ASJC Scopus subject areas

  • Medicine(all)

Cite this

@article{edbf947f70cc448c9854a4fbf7d00c6b,
title = "An application of Expert Network to clinical classification and MEDLINE indexing.",
abstract = "An effective and efficient learning method, Expert Network (ExpNet), is introduced in this paper. ExpNet predicts the related categories of an arbitrary text based on a search of its nearest neighbors in a set of training texts, and a reasoning from the expert-assigned categories of these neighbors. Evaluations in patient-record text classification and MEDLINE document indexing show a performance of ExpNet in recall and precision comparable to the Linear Least Squares Fit (LLSF) mapping method, and significantly better than other methods tested. We also observed that ExpNet is much more efficient than LLSF in computation. The total training and testing time on the patient-record text collection (6134 texts) was 4 minutes for ExpNet versus 96 minutes for LLSF; on the MEDLINE document collection (2344 documents), the total time was 15 minutes for ExpNet versus 4.6 hours for LLSF. It is evident in this study that human knowledge of text categorization can be statistically learned without expensive computation, and that ExpNet is such a solution.",
author = "Y. Yang and Christopher Chute",
year = "1994",
language = "English (US)",
pages = "157--161",
journal = "Proceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care",
issn = "0195-4210",

}

TY - JOUR

T1 - An application of Expert Network to clinical classification and MEDLINE indexing.

AU - Yang, Y.

AU - Chute, Christopher

PY - 1994

Y1 - 1994

N2 - An effective and efficient learning method, Expert Network (ExpNet), is introduced in this paper. ExpNet predicts the related categories of an arbitrary text based on a search of its nearest neighbors in a set of training texts, and a reasoning from the expert-assigned categories of these neighbors. Evaluations in patient-record text classification and MEDLINE document indexing show a performance of ExpNet in recall and precision comparable to the Linear Least Squares Fit (LLSF) mapping method, and significantly better than other methods tested. We also observed that ExpNet is much more efficient than LLSF in computation. The total training and testing time on the patient-record text collection (6134 texts) was 4 minutes for ExpNet versus 96 minutes for LLSF; on the MEDLINE document collection (2344 documents), the total time was 15 minutes for ExpNet versus 4.6 hours for LLSF. It is evident in this study that human knowledge of text categorization can be statistically learned without expensive computation, and that ExpNet is such a solution.

AB - An effective and efficient learning method, Expert Network (ExpNet), is introduced in this paper. ExpNet predicts the related categories of an arbitrary text based on a search of its nearest neighbors in a set of training texts, and a reasoning from the expert-assigned categories of these neighbors. Evaluations in patient-record text classification and MEDLINE document indexing show a performance of ExpNet in recall and precision comparable to the Linear Least Squares Fit (LLSF) mapping method, and significantly better than other methods tested. We also observed that ExpNet is much more efficient than LLSF in computation. The total training and testing time on the patient-record text collection (6134 texts) was 4 minutes for ExpNet versus 96 minutes for LLSF; on the MEDLINE document collection (2344 documents), the total time was 15 minutes for ExpNet versus 4.6 hours for LLSF. It is evident in this study that human knowledge of text categorization can be statistically learned without expensive computation, and that ExpNet is such a solution.

UR - http://www.scopus.com/inward/record.url?scp=0028310673&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0028310673&partnerID=8YFLogxK

M3 - Article

SP - 157

EP - 161

JO - Proceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care

JF - Proceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care

SN - 0195-4210

ER -