The MT@BZ project is a pilot study about Machine Translation at South Tyrolean Institutions.
The MT@BZ annotation guidelines are the first annotation guidelines for legal Italian-German machine translation quality assessment published so far for this language combination. Particularly, they cover the South Tyrolean German variety. They are based on version 1.3.3 of the Annotation Guidelines for English-Dutch Machine Translation Quality Assessment (https://www.lt3.ugent.be/publications/annotation-guidelines-for-english-dutch-machine-tr/).
The guidelines are structured into two main parts: accuracy errors and fluency errors. Accuracy errors lead to a target text that does not reflect the same information as the source text, while fluency errors relate to the correctness of the target language. In the guidelines, both categories are further split into subcategories, which describe in more details the types of mistakes made by the machine, for instance mistranslation, omission or foreign word. Each error category is described with a definition in English and further explained through one or more examples in the language pair Italian-German. The examples used in the guidelines come from the MT@BZ translation corpus. Finally, the guidelines also include specific instructions on how to annotate errors in WebAnno/INCEpTION and which sources to consult when assessing the correctness of a translation.
The MT@BZ annotation guidelines are available from the
Elena Chiocchetti, Flavia De Camillis
MT@BZ annotation guidelines is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
Any code or scripts in the repositories are licensed under their respective LICENSE files.