Google Translate is certainly the most famous online translation tool! A huge number of people use it every day. However, the fact is that all its users who better know the language they translate via this tool notice its shortcomings. It is not uncommon for Google Translate to make a translation that is meaningless! That is why it is constantly being worked on improving the performance of Google Translate, for the needs of which sometimes are engaged, guess who – translators!
If you thought that the group of people who designed Google Translate sat at a computer and entered words into its database in different languages - you were wrong! Google Translate works on the principle of linking words and syntactic constructions of one language with words and syntactic constructions of another, which it collects from over a trillion web pages located on Google. It links websites based on the technical similarities of the pages (e.g the same domain), and accepts those that it estimates to represent translations as a whole. So it collects words and expressions for its database.
The assumption was that access to a large amount of data would be better than smart access to a limited amount of data, that is, that in so many data the errors would cancel each other out. However, that did not turn out to be completely true. Errors are constantly occurring and everything goes to the point that Google Translate cannot provide authentic translations without the help of translators.
In addition to the large amount of data it receives from existing websites on Google, Google Translate also receives specific data from the databases of professional translation institutions such as the European Patent Office (EPO). It is interesting to note that the words obtained from these specific databases are not of any use when it comes to solving the problem of incorrect translations. The reason is the fact that Google Translate is most often used to translate words and expressions of spoken language and slang that are not found in such databases. This is another problem that only speakers of certain languages and translators can solve!
A special problem for Google Translate is represented by rare language pairs (e.g. Finnish – Zulu), for which there is very little text or no text at all on a large number of web pages. In that case, the translation goes through a “bridge” (the language in which there are translations for both languages). Most often it is English. This however increases the likelihood of errors and incorrect translations to appear. The question is: What does Google Translate do with the inaccurate data it collects? Does it use its “waste”?
When incorrect translations appear, Google Translate watermarks them so that the system recognizes them and avoids returning them to the database. On the other hand, the system is set to avoid web pages with translations in all languages existing in the Google Translate database, because it recognizes them as pages for the translation for which it was used. It automatically estimates that its translations are incorrect, and when it recognizes such pages, it rejects them.
And what does it do to prevent mistakes? It uses smart software solutions, such as automatic parsing that can improve word order in translations, or various neural networks for speech recognition. Nevertheless, even this does not guarantee the authenticity of translations obtained with the help of Google Translate! That is why it also relies on a proven good way of translating – translation by translators!
When it comes to standard language pairs, errors do not occur often. There is a large amount of data for them and the principle of spontaneous annulment of inaccurate data works well in this case. Help of people is needed when it comes to “less worldly” languages and specific language pairs that are rarely found among the data on the Internet. In that case, the speakers of these languages suggest and enter the correct translations in the defined Google Translate field (Help Improve Google Translate).
All in all, in the end it all comes down to the human factor! Translation is still the job most adequately performed by translators. Language is “living matter”, words and expressions are often translated differently depending on the context, and the context itself is most accurately recognized by man. Google Translate translations can be used for everyday purposes, but consultation with speakers or translators of the language from or into which it is translated is desirable. And when it comes to translations of important documents, they must be entrusted to professional translators and court interpreters. In the end, only they can guarantee the authenticity of their translations!
Source:
Greene, L. (2016). Everything you ever wanted to know about Google Translate, and finally got the chance to ask. TAUS Review # 3