Top 3 Products & Services
Dated: Jan. 18, 2013
Related CategoriesTechnological Advances
Google Translate and other free Internet services for translating enable users to access a text on any language, and the product quality ranges from impressive to not intelligent.
Internet connects the world, but most people are still separated with language barriers. That is why the mechanical translators like Google Translate, Microsoft Bing and Sistran are a bridge to other languages.
Franz Joseph Oh, a man that leads the Association for Machine Translation, says that over 200 million users worldwide click on Google Translate every month.
Still, there are no illusions about the challenges that represent making a logical text that is readable, and that has gone through mechanical translation. "If an internet page is on French, and you don't speak French, a mechanical translation is not as pleasant as human translation, but you can understand it." says Oh for DW.
IT IS A MATTER OF PROBABILITY
Goggles service for translation takes over the existing human translations on the internet and uses statistical algorithms to calculate the most probable match between two parallel texts in the given language pair. This method is based on a discovery from a team of IBM physicists in the last century. Goggles rival - the Microsoft service Bing, also uses a similar method.
A SMALL ERROR CAN COMPLETELY CHANGE THE MEANING
The translation service Sistran works on a different system. It uses the rules of linguistics. According to Alosa Burchard, the expert for mechanical translation in the German center for artificial intelligence in Berlin, one of the main advantages of this system is that it gives the user a larger result control. "You know what a system based on rules does, but you need experts to change a domain, that demands a new lexicon and a new way of forming sentences", explains Burchard for DW.
But that is not the case in systems based on probability, like Google or Bing. "Certain pages make good results, such a parliamentary debates, where there are tons of texts, on a large number of languages, that were translated by UN or EU translators." adds Burchard.
THE PROBLEM WITH GOOGLE TRANSLATE
Burchard notices that the quality of Goggles results is dropping drastically in certain areas where the availability of translated text for parallel comparison is sparse, areas like zoology or hunting. In that case mistakes are numerous.
Statistical mechanical translating depends on billions of translated sentences. The UN and EU are professional sources of high quality translations. But, on the internet there is a large number of human mistakes, and machines can't choose.
THE QUALITY OF TRANSLATION DEPENDS ON THE INPUTTED DATA
Google doesn't make a difference between human and system generated translations. "A lot of people puts their translation on Google Translate on the internet. The result of that is that our system learns from them, and we don't want that." says Oh.
His team works on algorithms to battle with bad mechanical translations. Also, Google fixes it's translations with a little help from the users, who translate texts on their maternal languages.
A new option in the system is that it asks to choose a better translation option. The representative of Google says: "It is good that people have common sense, that helps them to use Google translate, because even if it formulates something weird, they can still understand what it means".
Now that you've gotten free know-how on this topic, try to grow your skills even faster with online video training. Then finally, put these skills to the test and make a name for yourself by offering these skills to others by becoming a freelancer. There are literally 2000+ new projects that are posted every single freakin' day, no lie!