Automatic language translator - "Babylon Fish"

Will and when will we be able to build an automatic device to translate from language to language in real time, and what will be the effects of such a device that is available to anyone anywhere and immediately? What are the difficulties we have to solve before we can produce such a successful translator? All this and more, in the article before you

I'll start with a case that happened to me recently: I was sitting in the living room with my wife and her grandfather. The grandfather only speaks Spanish, and since all I know how to say in Spanish amounts more or less to "moi fun", the communication between us requires simultaneous translation by a third person. Suddenly an idea came to my mind. I opened the laptop with the wireless internet and entered the Google Translate site. I wrote simple sentences in English, and in an instant I received a translation in Spanish, which greatly facilitated the communication between us. So it's true that it's difficult to have an in-depth and fluent conversation this way, but it definitely allows for basic communication. What will the next generation of this interesting technology look like?

The term "Babel Fish" is taken from the wonderful book of Douglas Adams (zt'l) - "The Hitchhiker's Guide to the Galaxy". The book mentions, among other things, a tiny fish from another planet that is threaded into the ear, and because it is fed by brain waves, it creates an automatic translation between all the languages ​​in the universe, hence its name "Babylon Fish".

Well, we probably won't have to wait for alien fish that digest brain waves to be found. It is very possible that soon you will be able to find such an automatic language translator on your cell phone. Imagine a device (probably digital) that could translate a conversation in language A to language B in real time. I mean the situation where I speak Hebrew next to my cell phone, (or any other small device with a microphone), and immediately the words are played in English or French, or alternatively a text appears on the screen with the words I said in another language.

I will immediately get to the bottom line - such a high-level automatic translation device is not yet in sight. Despite tremendous progress in the field of language recognition by digital means that has been made in the last decades, we are still very far from being able to understand a conversation. In fact, even if we manage to analyze the conversation at a high level (recognize all the words spoken with an accuracy of 90% or more), we are even further away from being able to translate each word accurately according to the context of the conversation. The worse news is that without a significant and revolutionary breakthrough in computing, there are many obstacles in the way that will be nearly impossible to overcome. Those who want to understand more deeply why we are still far from the ability to analyze language, are invited to read the article I wrote on the subject and it is called "Is it possible to create real artificial intelligence".

Prof. David Harel, Dean of the Faculty of Mathematics and Computer Science at the Weizmann Institute, and winner of the Israel Prize for 2004, dedicated the last chapter of his book "The Computer Is Not Omnipotent" to the problem of understanding language - Natural language processing or NLP for short. He gives some amusing examples that illustrate the difficulty in understanding words: "The sentence: I don't know how to tell you," can easily be mistaken as saying: "I don't know how not to tell you." Similarly, the sentence: "I hate poetry", can sound like: "I hate a bad song" (p. 253).

He demonstrates the many problems involved in understanding the subtleties and semantic structure of a spoken language using the famous example that is both true and an urban legend according to which, during the Cold War, the US Department of Defense wanted to develop a computerized translator that could translate from Russian to English and vice versa. After a long time in which the scientists developed such a system, a senior officer arrived to examine the results of their work. He entered the sentence into the system: "The spirit is willing but the flesh is weak", and asked the computer to translate the sentence into Russian and the result back into English. The result was: "The vodka is strong but the meat is rotten". Prof. Harel summarizes the problem and writes: "We are not saying that it [understanding natural language] is impossible or hopeless, but only that it is much, much more difficult than it seems, and involves much, much more than meets the eye" (p. 256).

And despite all this at a basic level the technology already exists and you can find at least one commercial device that claims to perform a simultaneous translation of words from English to German, French or Spanish and even pronounce the translation aloud. This device is called "Universal Translator". I have never tried the device, but I find it hard to believe that it offers a particularly high capacity. The manufacturer's website says that they chose to translate a vocabulary that contains about 3,000 common expressions suitable for tourists. With such a limited amount of words, even if they are recognized in high percentages, it is clear that it is not possible to hold in-depth philosophical conversations. In addition to the device in question, there are many software in the world for translating sentences (especially words) that are able to translate in a way that will allow a general understanding of the translated document. A useful and good example of this is the Google translator that I already mentioned at the beginning of the article or the Alta Vista translator and the excellent AJAX translator that I use from time to time to translate from German, French and Chinese to English. Make no mistake - the translation is still at a very low level and far from the ability to translate complex sentences, but at the level of a single word or simple and short sentences, it provides reasonable results, but currently these programs require an Internet connection or at least the use of a real computer.

If and when we have such technology, what will it be used for? It seems that one of the most effective and common uses for such a device will be in the hands of tourists. Imagine a tourist who arrives in another country and does not have to break his teeth on the five words he learned in the local language. But there is a danger of a poor translation that you won't even be able to recognize immediately because you won't understand the language. It's good when it comes to an Israeli tourist who arrives in the US and asks a passerby "What is the time?", and the device will say something like: "? What is the hour", but what will happen if an American tourist says to the device: "? Where can I buy icecream" which may be Translated to an Israeli passerby as: "Where can I buy I'm screaming?"

With further developments and the expected technological progress, it can be assumed that in the next decade we will see more effective commercial attempts at such technology, and I believe that in the next twenty years we will be able to find in almost every cell phone (which will actually be a portable personal computer), a device that will work in a way that will allow basic but efficient and safe communication between Two speakers of different languages. Such a device could be used for military purposes in situations where soldiers need to communicate with a population that speaks a foreign language, workers at an international airport, embassies and other places where several languages ​​are spoken, and of course use for tourism and business needs.

(This article is taken from Amnon Carmel's blog which deals with futurism, technology, science and more)