BEGIN:VCALENDAR VERSION:2.0 PRODID:ILLC Website X-WR-TIMEZONE:Europe/Amsterdam BEGIN:VTIMEZONE TZID:Europe/Amsterdam X-LIC-LOCATION:Europe/Amsterdam BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:19700329T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:19701025T030000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:/NewsandEvents/Archives/2002/newsitem/222/7-Ju ne-2002-ILLC-talks-Wessel-Kraaij DTSTAMP:20020526T000000 SUMMARY:ILLC talks, Wessel Kraaij ATTENDEE;ROLE=Speaker:Wessel Kraaij (TNO Delft) DTSTART;TZID=Europe/Amsterdam:20020607T104500 DTEND;TZID=Europe/Amsterdam:20020607T000000 LOCATION:Diamantslijperij, Nieuwe Achtergracht 170 , room NP.101 DESCRIPTION:Abstract: Although Cross-Language Inf ormation Retrieval (CLIR) is available for more an d more language pairs, based on MT services, many languages are still not covered. An attractive alt ernative is to construct translation models based on parallel corpora mined from the web. Since thes e corpora are noisy, embedding translation models in the retrieval model is not without problems. Ho wever, experiments showed that retrieval effective ness can surpass MT systems in a CLIR evaluation t ask. X-ALT-DESC;FMTTYPE=text/html:\n\n
Abstract
:
\n Although Cross-Language Informati
on Retrieval (CLIR) is\n available for more
and more language pairs, based on MT\n ser
vices, many languages are still not covered. An a
ttractive\n alternative is to construct tra
nslation models based on\n parallel corpora
mined from the web. Since these corpora are\n
noisy, embedding translation models in the re
trieval model is\n not without problems. H
owever, experiments showed that\n retrieval
effectiveness can surpass MT systems in a CLIR\n
evaluation task.\n