Article written by Sarah Hickey.
Ever since it first appeared on the scene (see our article on the History of Machine Translation for the complete timeline), the use of machine translation (MT) in the language services industry has increased steadily. Once a cause for panic that sent many to believe that MT was the beginning of the end of the translation industry, today MT is just another tool in the toolbox and a normal part of the translation workflow at many language service providers (LSPs).
That being said, MT is not just one thing in itself. It rather represents a whole range of services LSPs offer to their clients. In this year’s survey for the Nimdzi 100 — our ranking of the largest LSPs in the world — we asked LSPs of all sizes what kinds of MT-related services they offer.
What type of MT-related service do you offer?
It should not come as a surprise that post-editing of MT content (MTPE) is by far the most commonly provide MT-related service. When MT is used, it is common practice (with exception) that the raw MT output is reviewed by a linguist to ensure quality and adherence to tone and style. That being said, some LSPs (37% of respondents from this survey) also offer raw MT as an option their clients can select. In such a case, as the name suggests, the machine output is not reviewed by a linguist, making it a much cheaper solution, though one that may only be suitable for low-value and low-risk content.
MT engine quality evaluation and customization are provided by 41% and 40% of respondents, respectively. Only 1% of LSPs in our survey provide MT platform services.
In addition to the types of MT-services on offer, we also asked LSPs to indicate the percentage of translation projects that typically leverage MT.
What percentage of your translation projects leverage machine translation?
The results show that for the LSP cohort who took our survey MT is, on average, leveraged in only about 25% of translation projects. While more than a third of respondents stated that they use MT in 1-10% of translation projects, this percentage falls steadily with 19.6% using it in 11-20% of projects, 15.2% in 21-30% of projects and finally only 8.7% of respondents leverage MT in more than 50% of translation projects.
If we consider that MT has become such an established piece of technology in the industry, at a first glance, these figures seem low. However, there are three factors to keep in mind when reading the results:
- Our survey is taking by LSPs who provide all kinds of language services, including interpreting, transcreation, dubbing, and subtitling — all services that either do not make any use of MT or to a minimal amount (for now).
- Some of the largest providers of translation services that are on our ranking, did not participate in our survey but provided us with their information through other channels.
- Our survey is not only answered by the largest players on the market but also small and medium-sized providers. Nimdzi’s research shows that, typically, MT adoption is lower for LSPs with smaller revenues.
Taking this into consideration, the survey results provide a good snapshot of the use of MT across the industry and for LSPs of all sizes.
This article was written by Sarah Hickey, Nimdzi's VP of Research. If you have any questions about interpreting technology, reach out to Sarah at [email protected].
