Our Project is to localize a real website by using transifex, a proxy translation platform. To avoid any copy rights issues, we decided to choose the website of a non-profit organization, and the final translated website will not be published. In this project, we use Plant with purpose as our source website. Transifex requires the users to choose the type of the projects, the files-based project is for localizers who have the website pages on files, and the Live Project option is for those who want to localize the website in real time through JavaScript. Transifex allows a team to administrate a project, which is convenient for a team to collaborate.
After adding the website URL to transifex, the system automatically generated a report, however, the word count report was incomplete because we had not yet make scans. Transifex counts detected strings only, and the word count report will be updated every time when the system detects new words or strings; moreover, the words showed on the report will be counted as using the “quota”, thus the users do not have to worry about wasting money on translating the words that they do not want to translate. This is different from Easyling (Easyling will count every word found in the first scan, so it is very hard to not exceed the word limit).
Translation memories can be imported and exported. This function will be particularly important if the translators need to convert the source language to tens of target languages. Each language will possess a unique report.
The most fascinating function in transifex is the live translation, allowing the user to apply a live preview of the translated website. When going into the live translation mode, the system will automatically run a scan through the current page, determining the strings to be translated. After the scan, users can approve the detected strings or exclude some of the strings, while approved strings will be ready for translation. All the translation jobs can be done in the editor. In this case, we used Google translation to machine translate all the approved strings. Marking the translated strings as “Reviewed” is all a user needs to do to finish the translation tasks. The live translation editor can be regarded as a small CAT tools, since we are doing the same thing as we will do in SDL Trados. No coding is needed, thus, even a translator without any relevant knowledge in programming can finish this on his own.
As mentioned before, transifex only run scans on current page, which means that following every link in the whole website and translating it separately is necessary. Links cannot be directly followed like what we do in a real website. Users need to click the button “follow the link” because a single click on the link is to edit the text, not to follow it. All of the translated pages can be chosen and reviewed by following the commands in the right tools bar.
The final job needed to be done is to publish the translated website. Transifex provides the project administrators with the JavaScript codes to show the translation. The project admins need to copy the codes and paste them onto every page that needed to be localized. Then, transifex also provide a staging server, which is staging.transifex.com. By using the staging server, localizers are able to test the project without publishing to the public.
Overall, transifex is a powerful and user-friendly proxy translation tools. It is easy to use, and the whole system is highly automated. It is definitely a very good choice for localization companies. However, its drawbacks are also apparent. The first one is the price. Transifex is much more expensive than its rivals, as its cheapest subscription is 179 dollar/month, with only 50K hosted words available. If a company does not have lots of website localization business, transifex is likely to be a bad choice. Second, although more than one user can localize the website at the same time, transifex sometimes crashes if two workers doing the jobs at the same time, especially if they are working on the translation for different pages, the auto detection may produce some bugs.