By definition, Translation Memory, or TM, is the computer memory that remembers previous translations of source and target language source strings. TM is essentially a database where bilingual text strings, called segments, are stored. Contrary to machine translation where complex algorithms are developed to make computer think like humans in processing natural languages, TM simply stores the bilingual pairs that a real human translator produces during translation. A TM file contains a table of source language and target language text strings in corresponding cells. During translation a TM tool will first looks up a text segment in a TM database to see if such a segment already exists. If so, the tool alerts the human translator that a match has been found. The translator can then verify the translation within the given context and select the given translation instead of translating the each segment from scratch.
Sometimes the source string and the target string are close, but not exact, matches. These matches are called "fuzzy matches". Though not word-for-word the same, these "fuzzy matches" are nonetheless helpful. Usually in this case, the human translator can still use the translation by only modifying the target text slightly and giving the translation a context appropriate, native-speaker sound instead of translating from each segment from scratch. Reusing previous translation segments is called TM leverage.

What are the benefits of TM?

Compared to conventional translation, TM leverage delivers three important benefits:
1. Reduced overall project cost.
2. Increased translation consistency.
3. Shorten translation cycle.

TM technology has matured substantially as the localization industry has evolved. Today, most TM tools are packed with sophisticated features that go far beyond just storing and leveraging translation pairs. Most TM tools today support multilingual terminology management, an in-context view of software user interface during translation and translation workflow management, to name just a few. TM is also very helpful for finding the appropriate terminology, or dealing with difficult sentence structures. TM is able to store all previous solutions that have been discovered and adopted in the past to assist the translator and save them time. Now the TM tool is even more powerful and is capable of helping translators and localization specialists by providing even more features. Functions such as autotext and automatic functions for checking formatting tags, numbers, and terminology make TM an even more powerful tool. While a full description of these advanced features is beyond the scope of the L10NWorks Website, there is widely available information from the different TM tool publishers and over the Internet.



For more information about CSOFT's translation memory , please click here.