Manual Quality Evaluation and Post-editing in Enhancing the Correctness of MateCat’s English-Polish Legal Translations
Edyta Źrałka · 2024
The reduced credibility of Machine Translation (MT) engines evoked the necessity of quality control and post-editing (PE). Some aiding instruments for evaluation were introduced (Quality Assessment Metrics - QAMs) to facilitate PE and make the process of MT valuable, subject to specified rules. The idea was not only to invent the criteria for metrics and evaluation performance but to record the proofread outcomes and make them repetitive. In such a case, the process of evaluation and PEwould be more effective and lead to better translation quality. A valuable contribution to that need was the creation of MateCat (Machine Translation Enhanced Computer Assisted Translation) tool, a combination of MT engine and CAT tool, enabling a user to translate automatically and edit MT results for better outcomes based on translation memories (TMs) and terminology databases. The outcomes depend on the tool’s parallel corpora and databases created by any individual user. Such linguistic data serve to improve the quality of subsequent translations, even if the language is specialised. The research aims to discover how the quality of legal texts translated via MateCat is enhanced based on quality assessment,PE, andthe creation of new databases.It also seeks to determine how the quality of the tool’s performance can be improved andproposes theoretical approaches to Translation Quality Assessment (TQA). Based on the research, it can be observed that introducing terminological corrections in MateCat translations results in consistently rendered terminology. Grammatical problems tend to be sustained due to fewer chances of contexts’ replicability. Keywords: Translation Quality Assessment, evaluation metrics, post-editing, CAT tools, translation memories.