QuEst - an open source tool for translation quality estimation
PASCAL2, Harvest program
SummaryThis project puts together many of the European leading researchers in quality estimation for machine translation to share their expertise in the field and: (i) jointly build an open source system for the problem; (ii) exploit new avenues for research in the field. Quality estimation for language output tasks such as machine translation is commonly framed as a supervised machine learning task. As such, it poses a number of challenging problems, such as the fact that it can exploit a large, diverse, but often noisy set of information sources, with a relatively small number of annotated data points, and it relies on human annotations that are often inconsistent due to the subjectivity of the task (quality judgements). Also, different applications for the quality predictions may benefit from different machine learning techniques, an aspect that has been mostly neglected so far. In addition to developing a state of the art, robust open source system to predict quality scores of machine translations aimed at different tasks, this project investigates ways of assessing the utility of the quality predictions in novel extrinsic tasks, such as self-learning of statistical machine translation systems.
SoftwareYou can download the software from here.