The development of the programme centres on statistical machine translation as opposed to expert human translators.
Previous programmes have used human language experts to program grammatical rules and dictionaries into computers. However, the machine logic approach feeds documents people have already translated into two languages; computers then discern patterns for future translations, working on a basis of statistical analysis.
Franz Och, head of Google's translation, admits that the process is not perfect, but that it is an improvement on previous machine translation concepts.
The programme may not be suited to accurate-critical document translations. However, where 'mostly correct translation' is acceptable, this new technology may signify a breakthrough.
The program is being developed to increase accuracy. Test translations have been carried out on certain documents from Arabic to English, and the results have not been perfect but are comprehendible. However, when translations have been tried from English to Russian, the translated documents have been said to be understandable but awkward.
The idea behind the machine logic is that, the more you feed into the system, the better it will be as it will have more data patterns on which to base its translation. Och has been feeding texts in Arabic and English, for example, using United Nations and European Union documents. The languages that are less frequently used, such as Africaans, will be more difficult to find and feed into the system. This means the overall system will be better at translating into some languages over others.
So far, the Google statistical machine translation system is offering translations in Arabic, Chinese and Russian to English. As yet, the tool is free; but if Google develop to a high accuracy standard they are likely to charge for its use in the future.
Limitations of the system have already been highlighted. Some argue that no matter how sophisticated statistical machine systems like this become, they will never overtake the accuracy of human translation. The system is certainly a valuable aid in the further understanding of documents in a foreign language, but not necessarily for producing accurate translations.
Eric Schmidt, Chariman of Google, believes that Google's translation system will be a breakthrough for societies in which language is a major barrier to communication.
