Machine translation is a hot topic among localization professionals. Translators and developers have sometimes spirited discussions about whether it can serve as a replacement for traditional manual translation. At the same time, almost all developers concede that no machine translator will be able to fully replace the human touch any time soon, especially for literary and creative texts.
How does machine translation work? Using a formal description of a human language, the machine translator analyzes content in that language, processes the information, and attempts to generate an equivalent phrase in a different language. This requires that programmers write extremely complicated algorithms for inferring the most likely meaning in the source language and how humans would tend to re-create it in the target language. Massive amounts of prior translations (corpora, consisting of source and translation texts created by humans) provide “seed material” for the software to learn what works best.
The benefit of these complicated algorithms is that the resulting machine translation allows instantly getting a “skimmable” rough translation for understanding the main ideas of even huge amounts of text. That said, the quality is not even comparable to that provided by a professional human translator. One way around this limitation is by giving the output text to a trained professional, who can edit it and improve the machine-generated results.
Machine translation is handy for visiting foreign sites and quickly getting the gist of the information posted there. It is also useful for quickly reviewing business or personal correspondence with people in other countries. Even professional translators can benefit from these tools. Since multiple translators might work on a large file simultaneously, a machine translator can help to ensure consistency.
With MT technology, editors can work more quickly and ensure that the same terms are used in the appropriate places. These tools enable translating documents quickly and to a high standard of quality, with consistent style.
Different MT systems have been developed for different industries and requirements. Some “engines” are made for translating from European languages into other ones, for example. In some cases, machine translation text is then post-edited by a human, with support for creating glossaries based on subject area:
MT tools are useful for professional translation agencies, who must handle larger and larger flows of manuals, guides and other technical documents into foreign languages.
Machine translation from SmartCAT can be used even without special dictionaries or glossaries, saving you time and effort. This approach is ideal for everyday correspondence and handling large amounts of interoffice communication. Today's tools can even process HTML documents without a hitch, since they include a special browser and other utilities. These machine translators are easy to use and available in both free and premium versions.
Whether or not you have existing glossaries and materials to help your machine translator “learn,” you can always add industry-specific and language-specific dictionaries. Smart MT systems are aware of morphology, which means that they understand all the forms taken by a word (so “go” is correctly understood as related to “went,” although “go” and “went” sound nothing alike).
Despite the amazing progress made, machine translation still has its drawbacks. It is limited to the rules and algorithms that have been programmed into it. Human language is incredibly diverse and sometimes vague, so misunderstandings are bound to crop up. One solution for making MT output more consistent and fluid, while fixing the biggest mistakes, is post-editing. This refers to when a human reviews MT output and manually corrects the text.
Several machine translation engines are available in SmartCAT, including ones from Google, Bing, and ABBYY. By signing up you instantly gain access to MT, glossary management, translation memory, vendor management, and all the latest tools for translation and localization. More information about SmartCAT features is available on the product site.