Skip to main navigation Skip to search Skip to main content

ICLEF 2003 at Maryland: Translation selection and document selection

  • Bonnie Dorr
  • , Daqing He
  • , Jun Luo
  • , Douglas W. Oard
  • , Richard Schwartz
  • , Jianqiang Wang
  • , David Zajic

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

6 Scopus citations

Abstract

Maryland performed two sets of experiments for the 2003 Cross-Language Evaluation Forum's interactive track, one focused on interactive selection of appropriate translations for query terms, the second focused on interactive selection of relevant documents. Translation selection was supported using possible synonyms discovered through back translation and two techniques for generating KeyWord In Context (KWIC) examples of usage. The results indicate that searchers typically achieved a similar search effectiveness using fewer query iterations when interactive translation selection was available. For document selection, a complete extract of the first 40 words of each news story was compared to a compressed extract generated using an automated parse-and-trim approach that approximates one way in which people can produce headlines. The results indicate that compressed "headlines" result in faster assessment, but with a 20% relative reduction in the Fα = 0.8 search effectiveness measure.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
EditorsCarol Peters, Julio Gonzalo, Martin Braschler, Michael Kluck
PublisherSpringer Verlag
Pages435-449
Number of pages15
ISBN (Print)9783540240174
DOIs
StatePublished - 2004

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3237

Fingerprint

Dive into the research topics of 'ICLEF 2003 at Maryland: Translation selection and document selection'. Together they form a unique fingerprint.

Cite this