How to stop machine translation taking its toll
Post-editing is easy when you know how
Every sector has a motto that ends up becoming famous, either because it is super powerful or just because it is repeated so much that it sticks, like “Beauty begins the moment you decide to be yourself” or “Content is king”. Even translation has one:
Automatic translation is here to stay
Whether we like it or not.
This is especially the case for localization; machine translation is combined with human translation to increase productivity, and lower rates, naturally. I have been working in the sector for 18 years and not a single client has offered me a machine translation without expecting reduced rates, so I know what I’m talking about. However, I digress; we’re not here to talk about rates today.
What is post-editing?
To keep it simple, post-editing is the process that involves adapting a text that has been translated with a machine translation system so that it sounds more natural in the target language and fulfils a given purpose. There are several degrees of post-editing depending on the text’s characteristics and the changes required by the client.
First, there’s light post-editing, where the message is more important than style, i.e. some mistakes are accepted if it does not affect the text’s meaning. For example, some spelling mistakes or sloppy grammatical constructions may pass, as long as the reader can understand the text.
Then, there’s full post-editing, which involves a more comprehensive post-editing process, where more attention is paid to both the message and the style. All spelling, typographical syntax and grammatical errors, etc., are corrected. This type of post-editing is more similar to editing, as its objective is to achieve an end text that reads like an original text in the target language.
Post-editing is a skill that needs to be learnt
This summer, I worked for a client who organised a post-editing course for its English-Spanish translators. First, I helped them classify the mistakes in the text translated with a machine translation engine (syntax, grammar, spelling, etc.). Then, I had to assess the texts that had been post-edited by the students, who I am going to call Oscar and Paula for the sake of this post.
When I was assessing Oscar and Paula’s work, I found two types of mistakes:
- Under correcting: they missed mistakes that needed correcting to ensure that the text fulfilled its purpose.
- Over correcting: they corrected things that improved the text but were not necessary to fulfil the text’s purpose, which also required extra effort on their part.
- Adding new mistakes: mistakes added by the translators.
What happens if we post-edit too much?
Translation is a very subjective process, which is why it is difficult to decide how much is too much when post-editing.
In my opinion, we need to be aware that even if we are doing a full post-edit, we’re not editing. If we were being expected to edit, the client would give us longer to do the task and would pay editing rates. If we were to spend the same amount of time post-editing as we would editing, we’re losing money.
But it’s not just a question of money.
The main problem with post-editing too much is that it’s tiring. When we’ve got a text that needs post-editing in front of us, we usually fall into the trap of taking it on with a lot of energy and changing more than is needed because we know that we have plenty of time and the end text will be better. As we progress, it becomes more and more tedious (I’m not going to lie, post-editing is not fun) and you’re more likely to make mistakes like Oscar and Paula did.
Just like other projects, there are machine translations that get the better of us for different reasons:
- The quality isn’t what we hoped it would be and we need to post-edit more than we’d thought.
- The text is too complicated (in which case, machine translation shouldn’t be used).
- We are tired and we can break our concentration easily.
If you are faced with a post-editing project, be sure about what the task involves and how to approach it. Don’t forget that it’s a machine translation that you need to improve to meet minimum requirements.
Personally, I don’t accept all machine translation jobs I receive. I think a machine translation also needs to fulfill some minimum requirements in order to work on it. Before accepting a machine translation job, I have a look at the content I have to post-edit and the client quality requirements. Machine translation is not suitable for all types of texts.
I know it’s controversial, so don’t hesitate to leave your opinion in the comments. All respectful opinions are more than welcome 😊.
See you for the next post!