I have a background in natural language processing and machine translation, and recently I've been interested in automatic evaluation of translations.

I've read a lot of literature from the machine translation world on approaches to evaluation. Most are statistical (e.g. n-grams), but a few use semantic or grammatical approaches.

I'm looking for inspiration from further afield and was wondering, is there a body of research within 'traditional' (non-computer) linguistics for evaluating translations? Can anyone recommend some keywords or papers to start with?

  • An excellent resource on the theory of translation from the point of view of changes in syntactic structure is provided by Lucien Tesniere (1959) in book E of Part I of his massive book "The Elements of Structural Syntax". It's in French, but an English translation will be available relatively soon. The chapter is pure syntactic theory, rich with numerous examples from a variety of languages. Commented May 11, 2014 at 19:24

1 Answer 1


In Translation Studies, the concept of 'translation evaluation' is a bit taboo, as it tends to consider that since there is not one single correct translation (esp. in literary texts), that means that there is no wrong translation as well. I personally tend to disagree.

However, you might want to look into the translator training sub-field of TS. That is bound to include some sort of evaluation. Also, there were some studies in the 1980s (the time TS was closer to Linguistic theory than ever) where translators were asked to describe the process of translating while they were doing so. Those descriptions, I think, sometimes also included self-evaluations of translation choices.

I know this doesn't exactly answer your question, but it might give you a few ideas. In case you are interested I can look up a few names/papers on the above for you.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.