文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Abstract Evaluation of French and English MeSH Indexing Systems with a Parallel Corpus
侵权投诉

Abstract Evaluation of French and English MeSH Indexing Systems with a Parallel Corpus

Objective: This paper presents the evaluation of two MeSH? indexing systems for French and English on a parallel corpus. Material and methods: We describe two automatic MeSH indexing systems- MTI for English, and MAIF for French. The French version of the

Evaluation of French and English MeSH Indexing Systems with a Parallel Corpus

Aurélie Névéola, b , James G. Morkc, Alan R. Aronsonc, and Stefan J. Darmonia, b

Laboratoire PSI - FRE 2645 CNRS - INSA de Rouen, France

aneveol@insa-rouen.fr

b

CISMeF - CHU de Rouen, 1, rue de Germont, 76031 Rouen, France

stefan.darmoni@chu-rouen.fr

c

National Library of Medicine, 8600 Rockville Pike, Bethesda, MD 20894, USA

{mork,alan}@nlm.nih.gov

tains 15 million MeSH-indexed resources in English. Since 1995, CISMeF2 (French acronym of Catalogue and Index of Abstract

Medical On-Line Resources) has been carrying on similar work on the most important resources of institutional health

information in French [1]. It currently contains more than

indexing systems for French and English on a parallel corpus.

14,000 resources selected for health professionals (e.g. evi-dence-based resources - practice guidelines & consensus con-ferences- and technical reports), medical students (e.g. lecture dexing systems - MTI for English, and MAIF for French. The

notes), and patients (e.g. patient education handouts). An av-French version of the evaluation resources has been manually

erage of 55 new resources are added each week. Indexing is a indexed with MeSH keyword/qualifier pairs. This professional

decisive step for the efficiency of information retrieval within indexing is used as our gold standard in the evaluation of

both the MEDLINE database and CISMeF catalogue, and it is both systems on keyword retrieval.

also one of the most time consuming tasks for the librarians.

This observation shows that it is necessary to develop auto-precision and recall (78% precision and 21% recall at rank 1,

matic tools to assist the human indexers in their work. Such

vs. 37%. precision and 6% recall for MAIF ). Moreover, the

systems have been developed for MeSH indexing in English

performance of both systems can be optimised by the break-as early as the 1980s [2]. More recently, MeSH indexing tools

age function used by the French system (MAIF), which selects

have also been available for French. This paper presents the

an adaptive number of descriptors for each resource indexed.

results of a comparative evaluation of two MeSH indexing systems, the Medical Text Indexer (MTI), which is used at the NLM to help with the indexing of English resources [9], and systems have features that can benefit each other.

the MesH Automatic Indexer for French (MAIF) which is currently developed within the CISMeF team. The evaluation

Keywords: aims at assessing both the systems performance and the com-plementarity of the methods implemented. Automatic Indexing, Internet, Medical Subject Headings, Par-allel Corpus. Materials and Methods

a

Introduction

The Internet has become a very ubiquitous source of informa-tion in numerous fields, including health. Several tools have been developed in order to meet the users' need to find pre-cisely what they are looking for in terms of health information among the numerous documents available online. With Pub-Med®1 and the MEDLINE® database, the U.S. National Li-brary of Medicine (NLM) was among the pioneers in medical information retrieval. Today, the MEDLINE database con-1

This section introduces the different elements involved in the evaluation, viz. the MeSH indexing systems developed both in the United States and France, the evaluation (parallel) corpus and the evaluation methods. NLM - Medical Text Indexer (MTI)

MTI results from the combination of two MeSH Indexing methods. These methods are a Natural Language Processing (NLP) approach based on MetaMap Indexing (MMI) and a

2

doc.guandang.net? doc.guandang.net

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email