The Categories of Translation panel discussion was held on March 9, 2023.
A recording of the discussion can be found here:
Description of the Panel Discussion
It was announced that this panel discussion would focus on categories of translation that can be part of the initial, pre-production discussion between a requester and a provider. Unsurprisingly, the moderator and panelists explored the entire translation process, not only the pre-production phase.
They did not reach consensus on categories of translation but afterwards agreed that the discussion should continue. FIT North America will probably host a follow-up panel discussion sometime in June. In the meantime, please watch the YouTube recording, which began just after the panelists introduced themselves. The names of the moderator and panelists are found at the end of the following description that was posted before the March 9th event. Although not emphasized explicitly during the panel discussion, the starting point in a requester-provider discussion is always a use case.
When a translation project is set up, the normal assumption is that (1) the translation should correspond to the source content, that is, the target-text reader’s interpretation should match the meaning intended by the source content author, and (2) that the target text should be fluent, that is, should read well in the target language.
For some use cases, such as triage or gisting, requirements for correspondence and fluency can be relaxed without undue risk. By clustering use cases into categories, according to required levels of correspondence and fluency, three “grades” (also called tiered requirement categories, TRCs) can be defined: high, medium, and low.
Product grade (TRC) can be included in a service level, along with turnaround speed, cost, workflow, and other specifications. The translation parameters in ASTM F575 and ISO 11669 were developed assuming that a “high-grade” translation is being requested. Only a subset of these parameters applies to low-grade (low-TRC) translation. [After this description was written, there has been a lively discussion about whether translation should be regarded as a product, a service, or both.]
While grades (TRCs) are defined by requirements, not method of production, low-end translation is typically produced using machine translation without bilingual review by a qualified human. This is called unedited machine translation or UMT. Something that is universally agreed on, so far as I am aware, is that UMT is only appropriate for selected use cases.
The kind of categories just described, defined by required levels of correspondence and fluency, can be called grades, but there has been a lot of push back on this term, at least in part because of the ambiguity with school grades. It would be correct to use the term Tiered Requirement Categories (TRCs), but this has not caught on.
A topic the panelists could discuss is a term other than “grades” or “TRCs” for the same notion. That is, if anyone has one to propose.
Another way to get away from the term “grades” is to focus on service levels. For a given item of source content, say a one-thousand-word document, a service level would include maximum turnaround time, agreed-on cost, and required minimum levels of correspondence and fluency.
Should professionals ever be asked to produce anything but a translation with the highest levels of correspondence and fluency? Probably not. Then what about use cases where the highest levels are not required? Are there any? If so, who or what should produce the translation in that case? These might be interesting questions for the panelists, unless they immediately reach consensus on the answers.
A service level should obviously be agreed on before production begins. During the post-production phase of a translation project, one might ask whether the agreed-on level of service was delivered. It is easy to determine whether the delivery deadline was met. Fluency can be evaluated to some extent by the end user, but level of correspondence can only be measured by a skilled bilingual with access to the specifications. Measuring correspondence is an important topic that is beyond the scope of this panel discussion. How to develop a full set of translation specifications is also for another day.
Each panelist will bring their own perspective on what should be included in a pre-production interaction between requester and provider. Should categories of translation be part of that interaction? If so, should they be defined by requirements for correspondence and fluency? If not, then how should they be defined? How should service levels be described?
Hopefully, by the end of the hour, everyone will have a better idea of where there is consensus and where further discussion is needed.
Dr. Don DePalma will be the moderator.
The panelists will be Donald Barabé, President of OTTIAQ, Dr. Arle Lommel, Senior Research Analyst at CSA, Dr. Pascale Elbaz, Chair of the FIT Research committee, and Steve Lank, who has been a translator and translation project manager and who has led the effort to update ASTM F2575, an international standard regarding translation.