Technical Translation

MACHINE TRANSLATION

In an age of automation, there is the constant drive to take on latest technologies to take the labor, expense, and time consumption out of those things that tend to result in a lot of labor, expense and time consumption.


Traditionally-speaking, translations/technical translations can be expensive and time-consuming. The cost of hiring translators is high (and inevitable). The processes of cultural research and redesigning documents can be slow. But there are certain technologies in use and in development that are reducing costs and enhancing efficiency.


One technology in particular is helping to automate translations with increasingly higher degrees of accuracy. Machine translation (MT) software is vital for industries with rapid and voluminous data turnover requiring frequent information updates (e.g. stock market and weather data). MT can also be effective for producing large, cohesive bodies of text that involve relatively little editing. 

Since MT continues to improve with advances in related software and hardware, increasingly more accurate, automated translations of complex content become more realistic. Thus, the burden of translating highly sophisticated technical documents can be made lighter through the use of MT software.

While MT can be divided into several categories; two particular methods stand out: statistical MT and example-based MT. 

Statistical MT uses statistical computer models to generate translations through an analysis of two bodies of text (corpora) in a database. These corpora consist of two different languages: the source language (language of origin) and the target language (language of the translation). The statistical models help determine the probability of linguistic matches when words are compared. The resulting translations tend to be more exacting than those from rule-based methods such as example-based MT; but, the results are generally inconsistent and require more specialized corpora and more and more sophisticated hardware.

The advancement of MT (namely statistical MT) depends on the willingness to support more collaborative research...

The example-based (rule-based) technique focuses on sample sentences or phrases in one language that are to be compared with sentences or phrases in another. The sentences are usually common, straightforward, and already translated. Blocks of words are translated through automated comparisons to similar blocks of words. While rule-based translations tend to be less fluent and accurate than the statistical translations, they tend to be more consistent and less reliant on specific, hard-to-produce corpora and don't require the volume or specialization of hardware.




Here is a good summary of MT with a synopsis of the rule-based vs. statistical MT argument...
http://www.systransoft.com/systran/corporate-profile/translation-technology/what-is-machine-translation