A translator’s job is a solitary experience. Whether hired as an employee or engaged as independent contractor, a translator works alone and in the moment. His cognitive effort, keystrokes, dictionaries, glossaries, dictation software and other tools, that is, his activities, transform one human language into another. He finds an assistive tool, like machine translation (MT) technology, beneficial or detrimental to the degree it contributes to or detract from his work.
Unfortunately, the traditional “Promised Land MT Paradigm” isn’t suited to support an individual translator as he or she works.
|Representative Set Averages for 31 Translators||GOOGLE|
|exact MT match (percent)||5.0%||34.0%|
|words per exact MT match (word count)||7||14|
|BLEU score – a likeness score (similar to TM fuzzy match) between MT and reference segments after-the-fact. Higher scores are better and 100 is an exact MT match|
exact MT match – MT segment exactly matches the reference human translated segment, AKA BLEU score 100 and 0 edit-distance (Levenstein)
MT technology has become a mainstream tool in the translation services market. Unfortunately, machine-translated segments are not created equal. Some are detrimental to a translator’s productivity. Therefore, a translator needs benchmarks that can predict, with a high degree of confidence, the likelihood that machine-translated segments from a given system will benefit his daily work productivity.
Our Study of Machine Translated Segment Pairs used two well-established quality estimation scores in a unique and innovative “Predictive Paradigm” to predict and describe the impact of machine-translated segments as an assistive tool on an individual translator’s daily work. The translator can use these benchmarks to maintain or extend his competitive advantage.
For LSPs: This study opens new opportunities to enhance your competitive advantage through new relationships with translators but this report does not address them. Subscribe to this blog or contact us to learn more. Do you have questions? Send us a message here to learn more.