Writing Web Content for Machine Translation

Global Translations provides a free Web Translation tool which you can use to add translation to your website or blog. With our free code, your Internet site will be available to readers in French, Italian, German, Spanish, Russian and Catalan. To get your free code, go to http://www.gts-translation.com/webtranslation.asp.

In recent years, Machine Translation (MT) technology has improved considerably, making it suitable for a wide range of translation tasks. However, quality is still lower than professional human translation. However, there are certain steps you can take to improve the quality of the machine translation. Here are some guidelines to writing content that will result in better machine translation quality.

Keep sentences and pages short

Avoid long and complex sentences. Write short sentences where possible, and cut up sentences that have several clauses. For example, turn the following sentence: “Introduced in 2002, the Acme widget sold over 100,000 units,” into “The Acme widget was introduced in 2002. The Acme widget has sold over 100,000 units.”

Try to keep web pages short too. Long pages take longer to translate-remember that the page is going through a MT server which processes the translation and more words take more time. Also, some MT servers have difficulties in handling very long pages.

Avoid use of ambiguous words, slang, idioms and jokes

When a word has more than one possible meaning, try to use the simplest word possible. The MT system may provide the incorrect translation where more than one option exists. For example, the word “spirits” means alcohol but is also used to indicate a temperament. When a MT server translates “David was in good spirits,” the machine translation may indicate that David is bathing in champagne. Another example: use “correct” rather than “right.” Another example: use “punctual” rather than “on time.”

Avoid the use of slang. In the USA, for example, “Holy Cow” is a term used to indicate astonishment at a rare feat. The MT engine does not know that however, and will render a literal translation. Jokes, idiomatic terms, country-specific sayings do not translate well so avoid using them.

Keep text separate from graphics

MT can not translate the text in graphics, such as buttons, banners and such. Texts in graphics will not be machine translated and will remain in the original language.

Grammar Guidelines

Use simple, direct sentences with basic grammatical construction. Ensure that the sentence structure is grammatically correct.

• Different languages make different distinctions among verb tenses. Tackle this problem by writing in the simple present tense whenever possible: Write "Dogs always chase cats" instead of "Dogs will always chase cats." Avoid the passive voice: "Exercise brings good health" works better than "Good health is brought about by exercise."
• Use the infinitive form of the verb rather than present participles because present participles do not always have equivalents in all languages. Write “Click here to select the icons and to view the images” instead of “Click here for selecting the icons and viewing the images.”
• Keep pronouns to a minimum. The meaning of pronouns can be lost after translation because different languages use different word orders and gender-specific languages may use different genders for certain objects. For example, in French ‘il’ could mean ‘he’ or ‘it’, so your subject may be unclear to a French reader. Replace pronouns with nouns wherever possible. For example, write “The book was a worldwide success: instead of “it was a worldwide success.”

Don’t be succinct

Don’t be stingy with connecting words such as “that,” “the,” “which” and “who.” These words describe the relationship between words in a paragraph and help make the MT more accurate. Instead of writing “make sure you turn on the power,” write “make sure that you turn on the power.”

Avoid acronyms

Most MT engines can not identify acronyms. If the acronym is in itself a word, such as FROST (Food Reserve on Space Trips), the MT will render the translation of the word in all caps. If you spell out the acronym in parentheses, the MT will translate the unabbreviated string but then the translation will not have the same first letters as the acronym.

If your material is technical and has many standard acronyms, consider a MT that has a customizable lexicon which supports tagging of acronyms and other buzz terms that may be best left untranslated.

Watch out for names that are words

If you put the following headline (appeared 5.8.06 on CNN.com): “Bush sends additional food aid to Darfur,” you get the following French MT: “Le buisson envoie à la nourriture supplémentaire de l'aide à Darfur.” The MT engine has no way of knowing that Bush means President Bush and not a shrub. The only way around this is use a MT engine that allows tagging of non-translatable strings or a customizable lexicon.

Garbage-In-Garbage-Out

If your source content is written poorly, has grammatical errors and spelling mistakes the MT will be equally poor, if not worse. Prepare your content meticulously.

Resources:

1. Preparing your Web site for machine translation, by Theresa A. O’Connell

2. Babel Not: Machine Translation for the Technical Communicator