iCLEF 2003 at Maryland: Translation Selection and Document Selection

TitleiCLEF 2003 at Maryland: Translation Selection and Document Selection
Publication TypeBook Chapters
Year of Publication2004
AuthorsDorr BJ, He D, Luo J, Oard D, Schwartz R, Wang J, Zajic D
EditorPeters C, Gonzalo J, Braschler M, Kluck M
Book TitleComparative Evaluation of Multilingual Information Access SystemsComparative Evaluation of Multilingual Information Access Systems
Series TitleLecture Notes in Computer Science
Volume3237
Pagination231 - 265
PublisherSpringer Berlin / Heidelberg
ISBN Number978-3-540-24017-4
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 F α = 0.8 search effectiveness measure.

URLhttp://dx.doi.org/10.1007/978-3-540-30222-3_42