If you want the short answer, it’s that there isn’t! There are easy ways of measuring certain aspects of a translation but no overall method of comparing the accuracy of one translated text with another. That’s because translations vary so much in length and complexity of content as well as delivery time. It’s like comparing fruit; it’s easier to compare apples with apples, rather than apples with pears. Even that is not strictly true as there are different sizes of apples, different maturity levels when they are actually eaten, let alone the fact that there are several different types of apples: granny smiths, braeburns, galas, red delicious and how about crab apples?
Getting back to translation and how easy it is to measure accuracy, there are several ways of measuring certain aspects of a completed translation piece, as long as they are not taken too seriously.
One way that has been used by some translation services is a quantitative measure. A translation is inspected for errors and a number assigned to each error based on its level of seriousness. For example, the most serious errors would be regarded as those that make the meaning of that section of the translation impossible to understand. Let’s say that gets a ‘3’. A less serious error that still may impact on the credibility of the translator or the translation service would merit a ‘2’. A minor error, say in grammar or spelling, would be given a ‘1’.
Theoretically, all that needs to be done is to tot up the points and divide by the number of words. This could be expressed as a ‘percentage error measure.’ It’s fairly crude as it doesn’t take into account the subject matter and the level of complexity of the content. It may be reasonable to expect that the more complex and specialised is the terminology used, the greater the frequency of errors.
However, as long as apples are being compared with apples in translation terms, i.e. translation tasks are compared with tasks of similar language content and complexity, then the measure is better than a subjective impression of accuracy which is surely the least accurate way of assessing accuracy.
The other issue that is not addressed by the method described above is whether the same mistake is being made throughout. This may be a simple misunderstanding of language and easily corrected yet would appear by to represent a ‘seriously inaccurate’ translation.
Despite the fact that there is no sure-fire way of measuring translation accuracy there are still ways that translation services anxious to monitor the quality of their work can do so, as long as they are not taken as gospel.