BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dynamic Topic Adaptation for Statistical Machine Translation - Eva
  Hasler (University of Cambridge)
DTSTART:20150306T133000Z
DTEND:20150306T143000Z
UID:TALK58003@talks.cam.ac.uk
CONTACT:Rogier van Dalen
DESCRIPTION:In recent years there has been an increased interest in domain
  adaptation techniques for statistical machine translation (SMT) to deal w
 ith the growing amount of data from different sources. Topic modelling tec
 hniques applied to SMT are closely related to the field of domain adaptati
 on but more flexible in modelling structure in between and beyond corpus b
 oundaries\, which are often arbitrary.\nIn this talk\, the main focus is o
 n dynamic translation model adaptation to texts of unknown origin\, which 
 is a typical scenario for an online MT engine translating web documents. W
 e introduce a new bilingual topic model for SMT that takes the entire docu
 ment context into account and directly estimates topic-dependent phrase tr
 anslation probabilities. We demonstrate the model's ability to improve ove
 r several domain adaptation baselines and provide evidence for the advanta
 ges of bilingual topic modelling for SMT over the more common monolingual 
 topic modelling.\nWe introduce another topic model for SMT which exploits 
 the distributional nature of phrase pair meaning by modelling topic distri
 butions over phrase pairs using their distributional profiles. Using this 
 model\, we explore combinations of local and global contextual information
  and demonstrate the usefulness of different levels of contextual informat
 ion. We investigate the relationship between domain adaptation and topic a
 daptation by combining both methods with automatic prediction of domain la
 bels at the document level.  
LOCATION:Department of Engineering - LR5
END:VEVENT
END:VCALENDAR