This page contains results of the Machine Translation Quality and Speed Surveys that were active during October and November, 2012. Surveys were run simultaneously in three languages: Spanish, French and Italian.
Multilizer thanks all the people who participated the surveys!
RESULT 1: Qualified Machine Translation Speeds Up the Translation Process
The Translation Speed Surveys included a sample translation material which consisted of the original text in English and the pre-produced material in the target language. The pre-produced material was created with MT-Qualifier and it contained automatically selected translations from several machine translation (MT) engines. The participants were asked to estimate how much time it would take to finish the translation with the pre-produced material as compared to translating from scratch.
Time is often the scarcest resource for a translator, and improvements in time efficiency are directly related to the overall productivity and profitability. The survey results show that translators can improve their productivity by using MT-Qualifier in their work. The large majority of repondents (56%) confirm that they can work faster and save time with qualified MT.
It is important to notice that when using qualified machine translation the work process will combine both translating from scratch and post-editing automatic translations. Not all translators are comfortable or familiar with post-editing, thus resulting why some respondents estimated that they could work faster when translating from scratch.
For those who have some experience in post-editing and are willing to optimize their translation processes qualified MT seems to offer a great opportunity.
→ Conclusion: Majority of translators can gain some or significant productivity advantage by using MT-Qualifier.
RESULT 2: Qualified MT Is Visibly Better than Raw MT
The Translation Quality Survey studied how qualified machine translation differs from raw, unprocessed machine translation. There was three pieces of text given to the respondents: the original text in English, raw machine translation in Spanish, and qualified machine translation in Spanish. The participants were asked to evaluate the quality of the both translations.
The results verify how qualified machine translation is able to automatically identify and filter the best available machine translation out of many sources. The quality of qualified MT was graded much higher than the quality of raw MT. Furthermore, the respondents stated that qualified MT didn’t contain any segments with totally incorrect or wrongly delivered information. In over 60% of cases MT-Qualifier conveys correctly all the core messages with none or only minor mistakes in secondary details.
The survey also highlighted that although any machine translation is rarely totally flawless, qualified MT is a good resource for post-editing purposes. Post-editing is most efficient when the automatic translation is good enough. Bad, low quality translations would only slow down translators’ work. MT-Qualifier is able to automatically select only those sentences that are fast to post-edit.
→ Conclusion: Translators can be sure that qualified MT contains only good translations for post-editing.