Context, Translation Quality, and Productivity
Accurate translation requires the translator to understand the context of the content they are working with and how it will be displayed. But that’s only the beginning of the problem. What is context exactly? And what is the most effective way to provide this context to a translator?
Context can take different shapes, depending on where the phrase you’re translating will appear. For a book, context would be the text surrounding the phrase you’re currently translating. For a website, context could be the entire content of the page on which the phrase is located. Similarly, for a mobile application, seeing the screen on which a phrase appears should provide information to the translator about what the context for this phrase is. It might even provide the translator with an idea of how much space their translation needs to fit in. For other content, like subtitles, seeing the video being subtitled is imperative to give an accurate translation.
Recently, a new crop of tools has surfaced promising translators the end of out-of-context translations, using a variety of rather clever techniques to do this. If you’re currently looking for a translation solution that puts context at the core of the translation process, there are several factors to keep in mind as you’re making your decision:
Being paid by the word, most professional translators prefer to translate using specialized software that will enhance their productivity: their bottom line is directly impacted by the amount of words they can translate in a given amount of time.
These tools provide a wealth of information to the translator about each piece of content that they need to translate, and allow the translator to navigate quickly from one piece of text to the next. There are two schools of thought here. Give context to the translator in their translation environment, or build a translation tool on top of the product, which allows the translator to translate in-context.
Because of their productivity boost and the amount of information translators need to provide a good translation, we think most translators prefer to translate using a dedicated tool. In this case, an in-context translation tool can be used by the translator as a second step to review or proof-read their work. Another solution is to make the feedback cycles between the translator and the developer as short as possible. This is what Transifex does with the Continuous Localization Process: a change made to a translation can be made immediately available in the product for translators to review.
Flexibility and control
Some of the tricks used by the in-context translation solutions are a fertile breeding ground for cross-site scripting attacks. How do you prevent malicious users from stealing your translator’s cookie and providing bad translations? If your localized content has to be delivered by a third-party, does their solution require you to route incoming requests, including those containing passwords, logins, credit-card information through their network?
Like most long lived engineering projects, you’re never really finished with internationalization. First you have to prepare your site for translation by extracting your content. Then you’ll realize that some of the source content needs to be split up into different translation units, because their context is different and requires different translations. You may even need to mirror some parts of your interface to support bi-directional text. Perhaps you will have to deal with different plural rules, or gender specific conjugations for some of the action words your site uses.
Some solutions provide you with an automated way to mark content up for translation. These solutions are usually based on a collection of regular-expressions or some more complex rules specific to the content you’re translating. Either way, identifying the content you want to translate is the first step to providing context. If you do not put care and thought into this process, you will get a lower quality end-result, because some content may be missing or needs to be endlessly modified to support all the languages you want to expand into.
Does the solution you’re evaluating support all the technologies you plan on using? Will it work for mobile applications, or HTML5 content? Will it work if your product is made of content from multiple sites like this page from l’Equipe, or gets embedded in another website, like those Disqus comments?
What about you? What are some of the other features and criteria you use to evaluate a localization platform?