Showing posts with label TMX. Show all posts
Showing posts with label TMX. Show all posts

Thursday, August 21, 2014

Maintaining TMs

What's the difference between 101% match and 100% match. Well, the main difference is a 101% match has been reviewed over time and is matched up not just as a previously translated sentence, but matches in context as well. This means the the translation is not only accurate but matches the style and context of the previous translation in the translation memory. LSPs and translators have faced this issue before. The match is 100%, but sometimes the context might be slightly incorrect or a word choice might need to be changed. For this reason, most LSPs would suggest a review of the 100% instead of ignoring it.  
Translation memory is a fabulous tool to enhance productivity, but if not used correctly becomes the repository for all sorts of old sentences, mistakes, and bad translation choices. Some TM's store multiple entires of the same translation without overriding the previous translation. This is an issue if multiple translators were used on a project. Now you have multiple enteries and inconsistent translations for the same source sentence. Translation Memories should be updated regularly especially after a client performs a review. 

One method to avoid inconsistencies is setting up a terminology database. This is one of the best tools to control quality and consistency of your translation. If terminology is analyzed and agreed in advance with your client and their in-country reviewers, it can only be the "right" terminology to use, thus cutting down on review time and enhancing quality, consistency, and brand messaging.

Downloading a TMX editor to clean up the inconsistencies, formatting tags, and miss-aligned segments is another method to help clean up the memory and ensure that you get better matches for future projects. This is important if you plan on using this translation memory to train an automated translation engine. 

The next thing that helps consistency is coming up with a style guide. Before starting a project, talk to your client, find out who their audience is, which style of writing they want to adopt. For many romance languages there is a great difference if you write in a “formal” manner or an “informal”. Then there is a difference in flavor, especially in Spanish. When a client says Spanish for Latin America, Ideally you want a translator who is a native of Argentina or Columbia. A translator from Mexico will have a different style and dialect that will not suit the Latin American audience. 

These methods will not only improve the quality of your translation, but it will reduce time and cost for you, your client, and the reviewers. 


Wednesday, February 25, 2009

How to Tango with Lingo


The ICD production room heats up every time there are files that are incompatible with SDLX™ or Trados™ . We have to find a work around, and it's usually more complicated or costly. Fortunately, the guys at MadCap Software™ decided to create Lingo that allows translation companies to work with files created by Flare™ and Blaze™ (MadCap's help file software). They were thinking ahead instead of waiting for SDLX or Trados to come up with a filter.

What it basically does is allow a translator to translate content from Flare into an SDLX environment without using SDLX. So you get the SDLX translation environment (left column source language segments/ right column target language segments) in Lingo, and it also allows you to create a translation memory for content reuse, and a TMX (Translation Memory eXchange) so you can import/export the file to SDLX or other translation tools. It also eliminates the problem of content transfer or extraction from a Madcap program into another software program. What it does not do is analyze the files for repeat content (saving you money), and build a memory from previously translated files.

One more thing, Lingo does not translate the content for you. Automated translations like Google translate are crappy and only work for very simple sentences or independent words. Technical terms and sentences with context value are usually inaccurate with automated translations. Therefore, if you are using Flare or Blaze, you can acquire Lingo, import the Flare or Blaze file, export your content as a TMX, and send the TMX file to the translation company. They'll analyze the file and translate the TMX. You get a translated TMX back, you import it in Lingo and you have a translated help file.

Well, thank you MadCap for thinking ahead, and reducing the stress for all of us in the translation and localization business.