Monthly Archives: January 2014

Machine Translation Used for Spanish Version of

By | Translations | No Comments

Turns out that the Spanish version of the new website relied on Machine Translation for translation.  We’ve blogged previously about the risks and costs associated with relying on machine translation.  Check out some of those for our take.

Alexandra Petri at the Washington Post does a great job of examining this latest round.  Read her take on it here:

Translation Memory Explained

By | Translations | No Comments

As we’ve discussed on this blog before translation is a complex process. Languages grow out of cultures and so even related tongues like English and French or German and Dutch can be separated by vast cultural gaps that a simple word-by-word literal translation does not convey. Anyone who has spent time trying to decipher poorly translated instructions on a toy or a bookshelf will tell you that quick and easy online translation programs often just don’t cut the cake.

Proper translation requires the work of a linguist who understands the nuance of both the languages involved. They need to be able to look at a sentence and not just substitute the words but to understand what the writer was trying to convey, and then express that in the grammar and syntax of the target language. It remains, however, a time consuming process. It takes many hours of work to rewrite a book or a manual, even for a trained and practiced translator. Given this problem, properly translating a work into multiple languages is often simply not cost effective. This is where the Translation Memory, or TM, has its time to shine.

Simply put, a Translation Memory is a program which handles the more repetitive parts of the translation work and allows the translator to spend their time and effort more efficiently. They work by first breaking the document, known as the source text, into contextual pieces, or segments. Segments can range from individual words to full sentences and once divided these segments are checked against a database of common phrases in the source language. The Translation Memory matches each of the segments in the source text to the database and rates them from 0% to 100% matched. When a 100% match occurs, the program will automatically substitute the segment in question with its established translation. Segments with a match rating between around 70% and 99% are tagged as fuzzy matches and are presented to the translator as suggested translations. Segments with a rating outside of that window are not considered accurate and are ignored by the program.

As well as searching for segments which match its database, the Translation Memory also looks for repeated phrases which it does not recognize. When the translator arrives at these phrases and translates them, the TM saves the translation into its database and uses it to translate the repeated phrase throughout the source text. In this way, the program begins to learn how to better translate the work in question and the process of translation accelerates with the Translation Memory doing more and more of the work involved. The translated segments are saved within the TM’s database for use in future works, ensuring that all the languages involved are kept up to date.

Translation Memories are highly beneficial for the translator and their client. By allowing the translator to work faster, they can deliver their finished produce to the client much sooner and, therefore, at a lower price than a translator who had worked unassisted. As such, Translation Memories offer to prospective clients a third option which is both cost effective for their business and assures the quality of a professional translation.

AZ World uses translation memories for all of our clients, and is proud to provide a more cost-effective product.  Request a quote at

Written by: Aaron Fenney