Machine translation services: using computer software to translate text
Machine translation is different from the process of using translators and a translation memory in that the process is completely automated.
It turns out that getting machines to produce high-quality translations is extremely difficult. Traditionally, machine translation and post-editing has resulted in poor quality output. Therefore most translation agencies, including TranslateMedia, do not work with translators who rely on machine translation. However, machine translation technology has improved drastically over the past few years.
As a result, a number of respected translators have gone on record to say that they find machine translation useful for speeding up the translation of simple documents and as a resource for stimulating ideas.
Customized machine translation
Customized machine translation engines can provide better results than general machine translation. This involves training a specific engine to handle your work as well as using human translators to edit the output, which trains the engine to improve future translations.
Here are some of the criteria that your work needs to meet in order to make this a realistic option for your business:
- At least 20,000 aligned translation segments (source and target text)
- A large body of mono-lingual reference data, to train the machine to adopt your brand’s style and tone
- A large body of bilingual reference data, to train the engine to improve its grammar
We have incorporated tools into our STREAM Translation Management System that allow you to train your own machine translation engine. Adopting this approach for a specific work stream within your organization may well provide savings over time, and by dictating a smaller domain for the machine translation to deal with, results can be drastically improved.
Rule-based & statistical machine translation
There are two main approaches to machine translation: rule-based, where the computer software attempts to model the rules of language; and statistical, where it tries to assume the translation from large amounts of previously translated text.
Statistical translation has improved substantially in recent years, and for some language combinations and types of text, the quality is now fairly reasonable. On very large projects, where it would not be economically feasible to translate everything using a professional translator, machine translation can be useful. However, post-editing by a professional human translator is highly recommended in order to improve quality.
We test Google’s machine translation software on a regular basis to see how the quality compares with other available tools. The hypothesis is that one day, Google’s quality will be sufficient that it can replace the translator in the translation process. In this case, the translator’s job may simply involve post-editing to prepare the document to the required quality standard.
However, we’re not there yet. All of our testing so far has shown that it is much more time-consuming correcting Google’s translations than employing a good translator to complete the translation from scratch. That being said, there are huge deviations from language pair to language pair, and depending on the type of text, some languages are better translated by machine translation tools than others.
In our tests, Google Translate has consistently performed better when translating English into Spanish and Italian than it has translating into French and German. It also translates general texts – simple communications and simple grammatical phrases – far better than more complex specialized texts. However, even in the best cases, we have found that the time required to correct the machine translated text means that it is more cost-effective to translate from scratch.
We are able to advise you on whether machine translation and post-editing services are suitable for your organization’s translation requirements. Contact us for more information, advice, a live demonstration, or a quote.