Translating websites with translation proxies

By: Marco Cevoli - Read Time: 8 minutes

Website localisation
Theoretical outlines of translation proxies
How does a translation proxy work?
Advantages of translation proxies
Disadvantages of translation proxies
Results of our survey
Webography
Translation proxy system providers

Translating websites with translation proxies is a guide resulting from research and a survey on the subject carried out in the last months of 2014. It explains how to translate websites without necessarily having to export and then import the content, thus avoiding cumbersome file transfers.

Website localisation

Translating and adapting websites is a complex operation, requiring close cooperation between at least four different profiles: client, web content creator, web developer and translator.

Broadly speaking, there are three traditional ways to localise a site:

direct translation within the back-end of the site;
external translation of previously exported content;
a combination of the two above methods.

This is a complex process involving at least five macro-phases, as illustrated below, plus a verification phase.

Stages of web localisation

A further complicating factor is the almost infinite number of different types of websites, so it is very difficult to define ‘formulas’ to fit all cases.

To simplify this process, client-based solutions have emerged that overturn the traditional approach. The term may indeed be reductionist, but for lack of a better choice, we will call these systems “translation proxies” or “localisation proxies”, using the term by which most vendors define their technology.

Theoretical outlines of translation proxies

What are these translation proxy systems? Before proceeding, a few theoretical points to understand how it works.

In computer science, a proxy is defined as a programme that stands between a client and a server, acting as an intermediary between the two, i.e. forwarding requests and responses from one to the other. The client connects to the proxy rather than directly to the server, sending it requests. The proxy in turn connects to the server and forwards the client’s request, then receives the server’s response and forwards it to the client.

In the case of a translation proxy, the proxy receives requests for translated pages and delivers them to the user, taking on the one hand the translated content from a database containing the original strings and their respective translations, and on the other hand the site structure from the original server, replacing the original texts of the site in real time with the corresponding translated texts.

How does a translation proxy work?

Scanning

It all starts by scanning the original site using a bot (which is a part of the system provided and is usually a simple script to be inserted into the site’s code). The bot examines the pages, saving the text strings it detects as translatable in a database.

Translation

The content of this database is then translated. This translation (and correction) phase can take place either within the translation proxy system, through an editor, or externally, by making the contents of the database available to the translators in the form of CSV, XLIFF or other files.

Importation

Translated content is then imported back into the database, either automatically if the translation takes place online, or by uploading the translated file if the translation takes place offline.

Translation proxy operation diagram

Illustration based on http://contenthere.net/2012/02/introduction-to-translation-proxies.html by Seth Gottlieb

This is the point at which the system acts as a true proxy. The translated version of the content is hosted on a proxy server and automatically replaces the content of the source language when a user accesses one of the site’s pages, intercepting the request.

The way it works is similar to that of advertisements, which are customised according to local settings. All this takes place in milliseconds and is invisible to the user, who simply views the site in the selected language.

Advantages of translation proxies

Certain features make these new systems very attractive:

The translatable contents of a site are automatically detected;
it’s possible to publish a translated site using zero lines of code, and zero developers;
also works with live and e-commerce sites;
opportunities are opened up for collaboration between translation proxy providers and translation agencies.

Within a classic web localisation process, the use of a translation proxy eliminates several steps, including:

the setup of a test server;
manual identification of contents;
the relevant extraction;
conversion into a “digestible” format by computer-aided translation systems;
subsequent import into the system.

Moreover, there is no need to interfere with dynamic content (articles, blog posts, user-created texts, etc.) on the one hand, or with the strings inserted in the interface of the template or framework used to build the site (buttons, system messages, etc.) on the other: everything is assembled by the bot and made translatable.

Another important point to consider: a translation proxy is a very practical way of making a site that was not designed to be localisable, thus freeing the client from the obligation to redo the whole site when they decide to go international.

We can therefore summarise the advantages of these systems as follows:

Advantages	for the client	for the translator	for the web developer
Avoids the need for complex text export/import procedures	✔	✔	✔
It even makes platforms that are not localisable		✔	✔
It exposes both “editorial” and template content to translators.	✔	✔
It centralises translations in an online platform	✔	✔
It simplifies managing updates	✔	✔	✔
The best systems offer contextualised text display		✔

Disadvantages of translation proxies

What are the disadvantages of translation proxies? Firstly, given the technology behind the system, translated sites are a replica of the original (the menu must be the same).

It is not possible, or rather, it goes beyond the logic of these systems, to have a complete site in one language and simplified versions for the others.

Technical issues aside (these limitations can be circumvented in some way), all commercial issues remain: The translations are stored on the servers of the service provider, to which you remain bound for the life of the translated site. The terms of service must be carefully examined in order to avoid misuse or critical confidentiality situations.

Another factor to take into account is the relationship between the translation proxy provider, the client and the translation provider.

Many translation proxies offer their services directly to translation companies (including white-label companies), who can then provide the localisation service to their clients. Clear supply contracts are needed to establish where the responsibilities of each lie.

Finally, since these are relatively new systems, it’s worth carefully assessing the reliability and robustness of the companies offering these services. It is important to avoid being exposed to changes in commercial policies or the disappearance of the supplier in the event of bankruptcy. However, this is an issue that arises whenever one entrusts a key business function to an external provider.

Results of our survey

In order to provide as complete an overview as possible of these relatively new systems, we asked 12 suppliers to complete a short online survey in October 2014. Of these, five responded. Some interesting data emerged from the results.

First of all, we found out that this sector is still young and that competition is fierce: four of the five companies have less than 10 employees, the youngest was established in 2013, the oldest in 2008.

Except for one open source system (SWeTE), all others use their own technology; both the open source solution and 3 of the other 4 are also offered as white labels.

On this subject, a brief note: a recent ruling by a U.S. court ruled that three of the MotionPoint company’s patents related to translation management systems and translation proxies were invalid because they infringed technology already patented by TransPerfect. This case sets a precedent, which could lead to a claim by TransPerfect against second-generation translation proxy system providers as well. For further details, see the analysis of this dispute by Donald DePalma.

In terms of functionality, the tendency is to offer very simplified systems, where translators must only deal with translation. In fact, all systems offer:

an online editor for translations, with a contextualised preview;
pre-translation by machine translation engines.

In addition to this, at the development level, all systems offer:

SEO functionality for translated sites;
full control over the CSS of the translated site;
a mechanism for replacing images on translated sites.

Only a few of them have connectors or plugins for the major CMSs (Joomla, WordPress and Drupal), whose functionality is still to be tested: integration with existing systems is an area which still requires work.

As far as agency-related functionalities are concerned, almost all systems offer the possibility to export strings as XLIFF (and to re-import them). Some (3 out of 5) also allow the use of translation memories, while only one also allows for a project glossary to be attached. Finally, only one system allows translations made in the online editor to be exported in TMX format.

All in all, the results show that these systems are mature enough to be used in the production environment, even if they are not always perfectly tailored to the needs of the professional translator or agency. Some are more advanced than others and the approach to the service is also different. Some offer ready-made systems, while others focus on extreme customisation.

Please visit the pages of the individual suppliers for more details.