Wiki Pages > Producing Information > Translation, Localization, and Globalization > Computer-Assisted Translation Basics
About Computer-Assisted Translation (CAT)Firms in the U.S. are increasing doing business on a global scale and need to produce technical documentation in multiple languages. A rough census of the Internet shows that the demand for multilingual content increases annually. It is reasonable to assume that the demand for multilingual technical communication increases likewise.
Annual Demand for Multilingual Internet Content
|Year||Number of Websites|
CAT and TerminologyCAT can compare source and translated documents for consistent use of terminology and then store these translations so that the same rules can be applied when aligning another source/translation pairs. The terminology management function gives the translator a means of automatically searching a database for terms appearing in a document to ensure that the correct source/target term combination has been used. The alignment function takes completed translations, divides source and target texts and segments, and compares them to determines how closely they match. Positive matches are used to build up a database of correct translations that can be reused. While terminology management can’t take context or localization into account, it can free the translator to focus on improving the usability of the translated material (Lionel Lim, personal communication, April, 2011)
CAT and Natural LanguageCAT parses text according to predefined linguistic rules that codify grammatical and stylistic requirements. Simplified Technical English and other versions of constrained English make rules-based translation easier because source texts written with a limited vocabulary leave fewer opportunities for mistranslation. One drawback: Rules-based methods generally sacrifice fluency in favor of predictable output. The statistical translation method relies on the global analysis of one language (e.g. English) and the same analysis of a second language (e.g. Spanish). The translation is based on the statistical likelihood that translated material will approximate the meaning of the source material. With the purely statistical method, there is no guarantee that identical source phrases will produce identical translated output.
|Rule-Based Processing||Statistical-Based Processing|
|Consistent and predictable quality||Unpredictable quality|
|Knows grammatical rules||Does not know grammar|
|Lack of fluency||Good fluency|
|Hard-to-handle exceptions to rules||Good for catching exceptions to rules|
|High development and customization costs||Low-cost development (Babelfish, Google Translate)|