There is some confusion between these two methods; Actually, the differences are so extreme from one another, so it is possible to relate to them as two opposites. In order to understand the severe drawbacks of the Machine Translation, let’s examine it from three aspects:
- As a response to the needs of sites and Enterprise.
- Potential market and business model.
- Development costs.
Machine Translation (as a Response to the Needs of Sites and Enterprise): There is no real Automatic Translation, but only a myth. The whole computer world has invested efforts in this field, but the technology still “works” only in Science Fiction stories. In the 80’s, the Japanese spent hundreds of millions of Dollars in this field (in the framework of the “5th Generation Computing” project), but failed. The Science Center of IBM in Haifa cancelled a similar project after almost 20 years of research. Friends of us, with a similar ambition, founded a cool startup (“SoftJet”) but failed too. And these are only three examples of hundredss of cancelled and failed projects.
What is the source of the difficulty? Contrary to the translation of separate words (e.g. Babylon.com), a human intelligence is needed here, to Understand the sentence, and with the lack of such a technology, there is no valuable translation engine available today. One of the ways that bored web surfers use to ammuse themselves, is to direct one of the web translation engines to a web document, and to be “tickled pink” as to the differences between the source and the result, and the absurd meaning.
There is a joke that illustrates this difficulty: When the CIA inaugurated its first English-Russian-English translation computer, one of the participants asked how one could check and see that the translation was the way it should be. The developers of the translation engine answered: “Type in an expression in English and see the translation”. And so the asker typed in the expression “Out of sight, out of mind”, and he received an answer in Russian. Someone else asked: “How do we know that this is the correct translation?”. He was answered: “Type the result back into the machine”. The translation was not long in coming: “Invisible idiot”. There is a similar joke with the translation of “the spirit is willing, but the flesh is weak” (Matthew 26) into “the vodka is good but the meat is rotten”.
It may happen even to humans. During the famous resolution of the UN at Lake Success, in 1947, to declare the establishment of a Jewish state in Israel, a reporter from the most popular Israeli newspaper translated the name of the city to Hebrew, and when asked to translate it back, he (mistakenly) translated it to “Lucky Pool”.
Although the best localization and translation bureaus are located in France, there are many jokes about the translation skills of French business, like the dress shop advertising “Dresses for street walking”, or the hotel’s notice “Please leave your values at the front desk”, or the lanudries’ “Ladies, leave your clothes here and spend the afternoon having a good time”. But the chance that a computer will make a mistake, is thousands of percents higher than a human translator.
Even separate words are difficult; Many English words have more than one meaning, and it is hard to guess how to translate “table” or “counter”.
Is it possible that an owner of a site will dare to risk and endanger the meaning of the contents in his site in the hands of only a machine? An e-commerce site may be a good idea: What can be better than letting the computer translate a disclaimer, or a guaranty, or a legal agreement regarding the buy, or a form with the details of the credit card. Or just the description of the products; After an automatic translation, the description will be so confusing that surfers will be too confused to sue the site…
We not only claim that no site owner will pay for a Machine Translation engine, but even that he may pay for a tool which will prevent other Machine Translation engines from translating his site (for example at the client side). There may be even a business potential, but this is not Our business.
Now more seriously: The only translation that truly satisfies site owners is human. The massive costs, even without counting the software, are a challenge, but (in a paradoxally way) may help the sales person to demand high prices for the software; When paying expensive salaries to human translators, the price of the software looks cheap. And this IS our business.
Potential Market and Business Model: So what is the target market of the automatic solutions?
Their target market is the surfers, that occasionally come across articles or items in an unknown language, and use those solutions to translate them. According to what we believe, this market is Bad. No sales, no customers, and no rules of the classic economy; Just the hope to make money out of banners. Contrary to sites and enterprise, private surfers can’t be charged for such services.
Moreover, this “market” is crowded, and there are already too many players competing. We can’t understand the logic of a company entering into this field.
Development Costs: Contrary to human-based translation systems, which only manage the process of the translation and let the human translators do the intelligent work, Machine Translation systems should understand the sentences. This requires an artificial intelligence and a very complex technology, even without trying to attack the difficulties mentioned above. A mediocre engine may cost millions to develop. Most of this technology must be re-developed again for any new language. And the meanings that are missed in any phase of translation (see the joke and the story above), force us to use only one phase, so no intermediate language is possible. It means that in order to support any combination of 10 languages, we will have to repeat the effort 90 (!) times, or to spend hundreds of millions.
Human Translation and Machine Translation may look similar from the first glance, but actually are two opposites. A real Machine Translation is only a myth, and even the existing solutions costs a lot to develop, and target a bad and crowded market of consumers who are not willing to pay for this service. On the other hand, there are sites and enterprises who are willing to pay a lot for a high-quality solution to manage the process of Human Translation. This solution is cheap to develop, and when based on a Proxy approach, is even unique and superior. This is our target market.