Quality Estimation Metrics and Analysis of Second Annotation Round and Error Profiles

This deliverable will evaluate the results of MT output from improved systems, comparing it to the results from D3.3, and the progress on fine-grained quality estimation models. It will also report on error profiles identified in the source and target languages, along with a recommendation for how to implement these in an automatic fashion. The report will address the significance of the factors identified and provide hypotheses for future testing in additional languages not covered in QT21. It will include an updated comprehensive database with annotated segments from both rounds of annotation.