What do you do when you have a continually changing stream of content that needs to be translated quickly, accurately and at a reasonable costs. Conventional LSPs aren’t really set up to handle this type of scenario, but fortunately there are highly automated services that can handle this type of process quite well.
Whether you are translating a company website or localizing an application, the translation work can typically be bundled into weekly or monthly projects that a conventional LSP (translation agency) can handle. This workflow breaks down when you have constantly changing content. Examples of this type of situation include:
- News services, where new stories need to be translated as soon as the source material is published.
- Vacation rentals and travel booking sites, which may have thousands of listings that are constantly being added or updated.
- Online storefronts, which may have thousands or tens of thousands of individual product listings, each of which has to be translated and kept up to date when the original item changes.
Even if you are using a translation management system (TMS) it is nearly impossible to keep pace with the changing source material (there are hacks you can build using continuous integration techniques, but its not easy or foolproof). Fortunately, there are several vendors that offer translation services via a web API.
These API based translation services enable you to fully automate the process of requesting and retrieving translations. To your system or application, they look a lot like a machine translation service, although the translations are done by professional translators and reviewers. The general workflow for your application looks something like this:
When a new source text is created or updated
- Make an API call to the translation service to request a translation for that text to the selected target languages, and to provide a callback URL for it to send the completed human translation to.
- The API immediately returns the translation (if that text has already been translated) or a machine translation that can be used as a placeholder.
- When the professional translation is done, the translation service makes an API call to the client’s web server to notify it that the translation has been completed.
- The client’s web server then makes an API call to the translation service to fetch the completed translation (sometimes steps 3 and 4 can be combined into a single callback request)
The details of this procedure vary slightly from vendor to vendor, but the general workflow is the same. These services generally return completed translations pretty quickly. I’ve seen turnaround times of just a few minutes with some providers, and in general most jobs are completed within 24 hours. The usual caveat applies that if you are using an uncommon language pair, it will often take longer due to translator scarcity.
The main weakness of these services is that the translators do not see the source material within the context of your application or surrounding material. This is not the vendor’s fault, it’s just a byproduct of this translation method. You can help translators by providing style guides, a term glossary, and instructions (screenshots are ideal, but generally impractical to generate). The results are generally good, much better than machine translation, though not as a good as translations that are proofed by testers who can see them in context.
The leading vendors in terms of fully automated translation include the following:
Gengo — one of the first, and probably the most technically astute translation agency in operation today. This is my go-to recommendation for software developers who need to incorporate translation into an automated process. Prices generally range from 5 to 14 cents per word depending on language and quality level. Note that Gengo was recently acquired by Lionbridge, but continues to operate as an independent business unit.
Transfluent — Transfluent, like Gengo, is an API based translation platform. They started out doing localization for gaming companies, and have since branched out into many industry verticals. They provide an additional layer of translation management compared to services like Gengo, so for example you can request translators who specialize in a particular field such as medical translation.
NOTE: this is not a complete list. I focused on translation providers that are focused on API based translations. Many LSPs will tell you they have an API which simply drops translation requests into their regular work queue. A well designed API will enable you to monitor job status, fetch statistics on delivery times and other metrics, and generally monitor progress in real-time or near real-time.
In general, these services (and others like them) are a good fit for situations where the translation needs to be fully automated. While they don’t deliver the highest possible quality (as they lack the ability to do in context review and testing), they generally provide good quality output, and at a reasonable price.