Introduction to Automatic Translation
An Introduction to the World of Automatic Translation
Language automation is amazing technology. We are overwhelmed and impressed with the progress by the companies whose software incorporated into Translate Your World. TYWI is the glue the enables these competing giants to function harmoniously as a team to produce this world’s first translator for across-language conversation.
Speech translation is comprised of 3 elements:
1) Recognizing what the person says in their own language and transcribing those words as text ("speech recognition")
2) Automatic translation of the transcribed text into another language ("automatic translation" or "machine translation")
3) Turning the text translation into a voice in the translation language (text-to-speech or "TTS")
#1 and #3 have been brilliantly handled by companies like Nuance, Google, Microsoft, Baidu and others. The problem has always been #2: automatic translation. But fabulous automatic translation is indeed possible. The major automobile companies produce hundreds of thousands of pages of manuals, instructions, and documentation in many languages through automation. The accuracy rate can be 98% to 99%, and require only a quick review by a professional (human) translator before distribution. How did they do it? They did not throw their car manuals at Google and say "Go!", rather they planned and organized to achieve phenomenal results. Translate Your World is trying to bring to you what these automobile companies have already learned – that automatic translation can be really good if handled carefully and in a planned manner.
Most people assume that translation is a swap of one word for another. So, "book" = "libro". The assumption that auto-translation is easy is a trap for companies and for developers. Businesses quickly learn that swapping one word for another is not sufficient. Considering the word "book", what if you wanted to "book a hotel room"? Oops! That is another meaning of "book".
Each profession or subject matter has its own vocabulary. “Coat” can be a jacket or it can be a coat of paint. “Application” can be software or a university admittance form. So, the first solution that TYWI provides is your own personal “translation dictionary”. You, your employees, or your listeners can enter preferred or corrected translations directly into TYWI. In that way, the correct translation of “book” will be inserted (pre-translated) before being sent to a translation software. This pushes out the important information to each translation software, and is a first major step toward personalizing translation to the subject of the conversation.
Now see what happens when we add another word: “red book” = “libro rojo”. Already the text is backwards in the translation. This next example clarifies word order for various languages as well as the differences in the spoken number (two).
|English||You have (2) messages in your mailbox|
|Japanese||Your mailbox in (2-ken) message there is.|
|Mandarin||In your mailbox (2-version#2- plus "tiao") message.|
|Thai||In your mailbox message (2) message you have.|
|Portuguese||You have (2-feminine) messages in your mailbox.|
|Arabic||You have twmessagi (blank) in your mailbox.|
|Hungarian||You (2-neutral) messages have.|
This is only a simple sentence, with few words. Yet the way the sentences are constructed are totally different. Importantly, as sentences become longer and add more phrases, the disparity becomes greater.
Thanks to the Giants
We are continually awed by what Microsoft, Google, Baidu, Systran, Yandex, SDL, and Asia-Online have achieved. What is also important to understand is that these companies are doing their best to please everyone. Yet each conversation is different, each subject matter uses different vocabulary. Perfect automatic translation for everyone is impossible to achieve at this time, so TYWI offers controls and enhancements through the TYWI software and its partners.
Differences Between Translation Software
All language translation "engines" are built upon different foundations. That is why Translate Your World offers various selections of automatic translation software, and we plan to add more soon. Translation software begins with a foundation one of the above grammar structures, and builds out from there.
One automatic translation software may be great for Spanish and Portuguese and Italian, but almost incomprehensible in Japanese or German. One correctly typed sentence with all the required accent marks may produce a perfect translation when passed through Google Translate, yet the same sentence without accent marks may produce utter nonsense. On the other hand Microsoft Bing may produce a correct translation for that sentence without accents, because Microsoft Bing often accommodates changes in orthography caused by texting, yet Microsoft in some other languages may be in the incorrect grammatical order or have dictionaries that are too small. Yandex may do a better job than any other with Russian, Greek, Polish, and Czech but not quite as strong on languages handled well by SDL, Google Translate and Microsoft Bing. Importantly, SDL especially and also Systran can be custom-tailored to your company and business as lightly or as powerfully as your contribution to the input, because these companies make personal translation engines available that permit companies to take control of their own results through feeding prior translations of documents and information into those systems.
Many of these software can be amazingly accurate – if handled seriously. And therein lies the issue. Most people expect machine translation to be perfect “out of the box”. People expect to speak any way they want, say anything they want, and expect the software to comprehend, decipher and translate flawlessly. Well, if it did, that would be a miracle, and we are in the technology business, not in the miracle business. 10 or 20 years from now, automatic translation will be a miracle, but why should we wait to start the world talking?
Personalized Translation Engines
Always use the TYWI translation dictionary regularly. For companies that are truly serious about automation accuracy, consider a personalized translation engine. Access to the personalized translation engines is available through TYWI. You can feed your documents, webpages, files, and even emails into your own translation software. These massive personalized translation software are what the automobile companies use to achieve such amazing results. But they are not plug-and-play, rather serious effort is required to feed these engines so that they have the background information to make correct vocabulary choices for the user. But once finished, your own personalized translation software is connected to TYWI and uses your preferred vocabulary, phrases, and content. You can also add preliminary document translation to the list of your capabilities with such an engine.
Looking to the Future
One of the major goals of Translate Your World is to combine users by profession, to share data, content, and vocabulary, so that groups by profession can take advantage of what has existed for a long time: personalized translation engines. These translation engines were not easily accessible due to cost and lack of competent advice. Translate Your World desires to give the world the ability to create profession-specific automatic translation engines in order to lower the cost of personalized translation software by sharing across users, and achieve the accuracy for which the world is clamoring.