In April of 2006, our translation agency was approached by Vagas, a Brazilian company that operates a site offering employment and career services (vagas.com). I had met the company’s founder sometime around 1985 when we both worked at a Brazilian factory that assembled computers, and I knew that he was a professional deeply interested in technology and dedicated to quality. For this reason, I went to the meeting fully prepared for some type of challenge.

Basically, the problem presented was the following: the site has a great volume of text that was all conceptualized and written in Portuguese, and the company was prepared to go international. In order to do so, they needed to make their site available in English and Spanish. What’s more, because they were in an extremely dynamic business sector, the content was subject to frequent updates, and the translated sites had to reflect these updates as soon as possible.

We decided to propose a solution based on a translation memory tool.

In a typical job using this type of tool, when requesting a translation of a system that has just been updated, the client simply sends the updated files in the original language to the translation agency. Then the agency performs an operation called file analysis, in which the translation memory tool compares the files to be translated with the existing translation memory and identifies the new text, which may be spread across several different files. This new text is translated by the professionals at the agency and the same tool inserts the text in the files, merging it with material that was already in the memory. The agency then delivers the complete updated set of the translated files to the client.

Nevertheless, in this case, the number of files was very large (in the thousands) and the updates, executed on a higher level of the structure, may affect any of these files. The standard solution for each update then, would be for the client to send us all of the files, and for us to analyze the material, translate the parts that have been altered, which may be spread across a wide range of files, and generate a new complete set of translated files to return to the client. If this was to be done on a weekly or monthly basis, it would require us to be responsible for constantly handling a large volume of material, which would have to be compensated in some way, thus increasing costs and creating a greater possibility for errors.

Which led me to ask: why does this procedure of handling the entire set of files need to be done by us? After all, the company is a producer of software. And for a company that has the human resources needed for the creation of software, operating a tool designed for translators should be akin to asking an Olympic cyclist to make a local bike delivery.

Based on this, we presented them with the following proposal:

• They would buy a license for the translation memory and learn how to use the module that performs the file analysis;
• Each time they wanted to translate their updates, they would perform the analysis of the set of files themselves and export only the unknown segments, meaning they would export just the text that is not in the memory to a single file. This exported file would then be the only thing they would send us, together with the analysis that was already done, so that the volume of the service would already have been pre-calculated by the client, facilitating our billing process;
• We would then just translate the unknown segments and send the same file back to them, translated;
• Then they could import the translated material into the total set of files, output the updated files, and implement the system update.
  The solution, then, optimized the costs involved, making use of the best resources of each of the two companies: we would handle the linguistic aspects, which was to do the actual translation, while they would contribute with the software operations.

The proposal was accepted.

− And then? How did it work out?

Perfectly. After that proposal, the translation memory tool has changed from Trados to a more modern application, SDL Studio.   So far, after 14 years, the solution continues to work without a hitch. Vagas sends us the updates in an extremely simple format, in files exported by SDL Studio, receives the translations, imports them to their system, outputs the results to its own files, and the translated site is constantly updated at optimal cost.

And since I’m on the subject, I might add that today, the 17 of August de 2020, we are sending Vagas another file with updates. With this delivery, we’ve made 97 updates since 2006.

This text was approved by: VAGAS

Marcos Chiquetto

Founder and director – LatinLanguages

< Back