Lecture

This PhD dissertation focus on the development of new methods for word-level
quality estimation in the two most important translation technologies: machine
translation and translation memories. The research carried out builds on the
use of external sources of bilingual information, that is, any resources able
to produce translations at the sub-segment level, to develop these techniques;
this allows, not only to obtain a collection of methods for word-level
quality estimation that are robust to any kind of sources of information,
but also to find new applications for existing resources. In order to enable
the application of the new methods developed for under-resourced language
pairs, part of the research also focus on the automatic creation of sources
of bilingual information from multilingual websites.