All the talk about machine translation (MT) has led to a common misunderstanding about just how good, and useful, MT is. Specifically, many people are unaware of the large part that training of the MT software and post-editing play in in the process.
You may read popular articles about the huge strides machine translation has made in recent years. You may have heard that several large companies use machine translation regularly to translate large volumes of technical materials. This is all true, but this does not mean that companies simply run their documents through Google, Babelfish or another online automated translation tool and their work is done (as many people have tried to do with their materials before calling us for help).
Large companies do use MT to speed up and ultimately save money on the translation of large volumes of repetitive material ; technical material like user manuals is the most common. Doing this, though, requires significant initial investments of time and money, as well as ongoing expenses for maintenance and editing.
- Changes in authoring process: This means the adoption of standardized style guides, term glossaries, and some sort of controlled language, such as “simplifed english.” This ensures that the writing is simple, unambiguous and consistent, so that it is suitable for translation by an automated system that needs to have a set of rules to be effective.
- Choice of software: Not all translation software works equally well for all languages. Research must be done and the software or online service purchased – at significant expense.
- MT training: Then the software has to be “trained” on the specific types of materials that it will be handling. This basically involves “feeding” the software preexisting texts with their translations so that it had a basis for recognizing future input. This training in ongoing. Errors in translation can be “fixed” in the MT database, avoiding similar errors in the future.
- Post-editing: Using MT doesn’t mean these companies dispense with linguists. The final step after using the prepared MT software to “pre-translate” the materials, is for a linguist to do post-editing. The extent of the post-editing depends upon what the final quality is that the client desires. At the least, the post-editor insures that the translation has no actual mistranslations and that it is comprehensible – though it may not read completely smoothly.
So, before you decide that Google Translate can solve all your problems, you should realize that, even to ensure that a product of MT is going to be understandable to the recipient, you should at least have it checked by a professional translator. If you need a high quality translation, as for marketing materials, business proposals, etc., you still need human translation to get it right.
If you’re wondering about when to use MT, take a look at our new publications, “When to Consider Machine Translation“.