Finding your language wallah

The Machine Translation Archive provides a useful resource to learn about the effort put into cracking the information age’s Tower of Babel problem.

I have licensed translation tools from Systran, which bills itself as “the leading supplier of language translation software.” I used Systran as my translation wallah (an Anglo-Indian word for a person in charge of a task). The company offers a range of translation products for commercial and government use. A home office version of the translation system is available in language pair packs—for example, English to German and German to English. Systran supports 52 languages. If you want to experiment with the system, you can purchase a Web translator for about $50 or a home office suite for $800. The small business and enterprise versions are priced based on the language pairs you want to translate and the components, such as workflow components, you require to solve your organization’s language problems.

My experience with Systran has been largely positive. I have a working knowledge of several languages. I can figure out an unclear passage by looking at the source document and the Systran version. When I need verification, I have to turn to a native speaker of the source language. Over the years, my need to tap a native speaker has decreased. In my opinion, the quality of machine translation systems has improved incrementally over the last five years. Most outputs are “good enough” for the type of work I perform.

Words of warning

When using machine translation systems to convert a document in English to a version in German, for example, close attention to word choice and sentence structure pays dividends. I get better results when I submit text written in short, declarative sentences. We have learned to avoid idioms, potential confusing homonyms like “wait” and “weight,” and neologisms like the awful coinage “webinar.” Touching up is necessary.

Commercial machine translation software offers a wide range of options. For example, for me to add a double byte language to my basic Systran system, I must purchase additional language packs. A “language pack” is essentially the knowledgebase that makes it possible for the Systran software to convert a language. For example, the English World Pack, which supports Chinese, Dutch, French, German, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish and Swedish, costs about $1,000. The company offers a discount for purchases of more than five licenses.

Systran, however, is a company that has had to deal with a somewhat unexpected disruption in the world of machine translation. In the last fiscal year for Systran, the company’s revenues suffered a decline. In the most recent financial reports available for the French company, the firm continues to suffer in the present economic climate.

Translation services from Google

I am reluctant to make a cause-and-effect connection between the deterioration of Systran’s revenues and the increased capabilities of Google’s translation services, but Google’s expanding machine translation capabilities may be a contributing factor.

Google’s translation services and its translation system are free as I write this. Try the Google system by pointing your browser to http://www.translate.google.com. The first thing to note about the page is that unlike Yahoo’s Babel Fish, Google places no advertising between you and the translation system. Second, the user can provide a URL, paste text into the translation box or upload a document. Fifty-two languages are supported. In the tests I ran, I did not encounter any document length limits.

The system supports multilingual search. You can enter a phrase in your native language. The Google system will translate your query into the target languages supported by the system. The search results will be returned regardless of the language of the page indexed by Google. You can translate pages in different languages using the Google Translate function. I saw a demonstration of a search system developed by Pertimm, located near Paris, that supported multilingual queries. Google’s implementation is a major advance in search functionality in my opinion.

There is a Translate option on the Google Docs menu bar as well. When I tested the service, I logged in and created a new document. I typed sample text into the document window. I then clicked on the Tools menu option and selected Translate. The translated document appears in the document window. The translate window offers two new options. You can replace the original document with the translation and copy the translation to a new document. One slick feature for me was that the translation preserves the formatting of the original document.