If we consider that the reason that our clients seek translation services in the first place is because they lack the ability to communicate in a given language, it naturally follows that the client is not generally in a position to judge how accurately a translation conveys the ideas and intentions of a source document. Economic realities pressure both PMs and translators to cut costs and/or increase capacity, which often leads to assigning and/or accepting more work than is prudent within a given timeframe. Poor quality translation can lead to problems during the proofreading phase, cost overruns, missed deadlines, and lost clients. Having clearly defined quality standards in place, as well as mechanisms for verifying compliance with them, will go a long way towards preventing the quality of translation work product from slipping in favor of increased production once business pressures come into play.
Why should the quality assurance process end once a translator is hired?There are several quality standards that regulate quality in the translation industry, including provider-based standards such as the American Translators Association Certification, process-based standards such as ISO 9001, EN 15038:2006, and ASTM F2575-06, and product-based standards such as the LISA QA Model. Any of these standards have the potential to play an important role in a translation quality assurance policy, yet while virtually all Language Service Providers (LSPs) have mechanisms for vetting potential translators, many do not continue to verify the quality of the translation being provided to them after the initial hiring process. I will suggest that LSPs can define and regulate vendor compliance with a quality standard in two ways: First, LSPs should set a quality standard by establishing their own definition of “acceptable quality” translation. Second, LSPs should conduct standardized quality evaluations to verify that the quality of the translation work product that it outsources is in compliance with that standard.
Setting a quality standard
First, an LSP must define acceptable quality translation. This can present a challenge, as “quality” can mean very different things in different contexts. For example, in a scientific context, a literal rendition that preserves the exact terminology and grammatical structures of the source text may be desirable, whereas in a literary context, quality translation is expected to include creative liberties. Despite these challenges, it is useful to define quality translation as the necessary first step in creating a translation quality assurance policy. Such a definition must be general enough to encompass a wide range of fields yet be specific enough to provide useful metrics. By way of an example, below is a definition of acceptable quality translation that I provide on my website as part of my terms and conditions for accepting proofreading jobs.
Acceptable quality translation means properly redacted and naturally phrased target language text using terminology appropriate to the subject matter that is essentially free from grammar errors or mistranslations, but in no case contains more than 2-3 minor error or 1 major error per 1,000 words, and is entirely free from machine translation, inconsistent translation, and nonsense translation in general.
Regulating compliance with a quality standard
Once an LSP has a specific definition of acceptable quality translation, translation work product can be rated based on whether or not it meets this definition. The most practical way to do this without impacting budgets or working procedures is to create a simple evaluation form for proofreaders to fill out. A quantitative quality evaluation, such one based on the LISA QA Model, could assign a value to each error by severity, and then use a formula based on the number of error points received divided by the number of words to assign each translation a quality score. The advantage of a quantitative quality evaluation is that its mechanical nature removes proofreader bias.
Below is an example of quantitative quality evaluation based on the LISA QA Model:
However, there may be some cases, such as literary translation, marketing copy, or other creative texts, when a mechanical error-counting method would not be the most suitable approach for rating translation quality. In these cases, a qualitative approach that allows the proofreader some autonomy to verify and comment on the essential elements of a translation would be useful.
Below is an example of qualitative quality evaluation that has been designed to allow the proofreader to take into account a linguist’s creative efforts.
From a linguist’s perspective, I would like to see many LSPs raise the bar for quality and document the quality of translation work product delivered back by each linguist for several reasons. First, having a linguist’s documented quality track record on hand would increase linguist accountability. Those linguists who consistently deliver back the highest quality translation could be identified and prioritized during the job assignment process while problem linguists could be warned and eventually removed from an LSP’s database. Second, the proposed quantitative or qualitative quality evaluation forms could serve as a training tool if returned back to the translator along with the proofread files (with tracked changes). This would provide the linguist the opportunity to learn from his or her mistakes and to adjust his or her style or terminology preferences to the client’s taste. Finally, defining translation quality and regulating compliance with a quality standard would promote fair competition since increased accountability would prevent inexperienced professionals from underbidding qualified linguists only to deliver back inferior quality translation work product for proofreading. Translation buyers, LSPs, proofreaders, and qualified linguists should welcome more stringent translation quality controls.