Friday, December 7, 2012

Improving the Computational Power of Google Translator Tool

Improving the Computational Power of Google Translator Tool
Using the tremendous computing power smoothing of the translation. At a meeting held at Google in 2004, the subject focus on a company's e-mail received from a fan in South Korea. Sergey Brin, which is one of the two founders of Google company, running the message through an automatic translation service that the company has patented.

The message read Google was a favorite search engine, but the translation says: "Sliced ​​raw fish will put horseshoes. Google green onion thing!"

Mr. Brin said Google should have done better. Six years later, a free translation service Google handles 52 languages, more than any other similar system, and people use it hundreds of millions of times a week to translate Web pages and other text.

"What you see on the Google translation of a work of art" in computer translations that are not limited to a particular subject area, said Alon Lavie, an associate research professor in the Language Technologies Institute at Carnegie Mellon University, told The New York Times.

Google's efforts to expand more than just a search engine has produced many successes. Digital books have been hung in the court, and the introduction of social networking, Buzz, increasing fears of privacy. Examples show that Google could one move when he tries to challenge business traditions and cultural conventions.

But Google's quick rise to the level of the upper echelons of the translation business is a reminder of what would happen if Google releases a tremendous computing power on complex problems.

Network of data centers for Web search dibangunya now could be the world's largest computer combined. Google uses the machine to push the limits of translation technology. For example, about two months ago he said that he was working combine its translation tool with image analysis to translate the writing in the picture.

"Machine translation is one of the best examples that shows Google's strategic vision," said Tim O'Reilly is the founder and head of the technology publisher O'Reilly Media. "It's not something that is taken very seriously by anyone. But Google understands something about data that no one else knows, and he's willing to make the investments necessary to tackle these complex issues ahead of the market."

Creating a translation machine has long been seen as one of the most difficult challenges in the field of artificial intelligence. Over the decades, computer scientists tried using a rules-based approach that teaches the computer the linguistic rules of two languages ​​and giving it the necessary dictionaries.

But in the mid-1990s, researchers began favoring a statistical approach. They found that if they tell the computer thousands or millions of paragraphs along with the translation of the human, the computer can learn to make accurate guesses about how to translate new texts.

It turns out that this technique requires very large data and computing power, it is suitable to the circumstances of Google.

"Our infrastructure is in accordance with this," said Vic Gundotra who is vice president of engineering at Google. "We can do the approaches could not even dream of the other party."

Automatic translation systems are far from perfect, and even Google will not make human translators out of work in the near future. Experts say it is difficult for a computer to break a sentence into parts and then translate and integrate them back.

But Google services good enough to convey the essence of a news article, and it has become an instant source of translations for millions of people. "If you want a rough translation of fast food, it's a place to visit. Said Philip Resnik which is a machine translation expert and an associate professor of linguistics at the University of Maryland, College Park.

Like its competitors in the field, most notably Microsoft and IBM, Google has incorporated interpreter engine with transcripts of the opening of the United Nations are translated by humans into six languages, and of the European Parliament which was translated into 23 languages. These basic materials were used to train the system with the most common languages.

But Google has been exploring Web script, as well as data from the book scanning project and other sources, to move beyond those languages. For more obscure languages, he brings a "toolbox" that helps the users in the translation and then add these texts into their databases.

Google Offers could hurt sales of software from companies such as IBM But the automatic translation is unlikely to ever be producing a lot of money, at least not of the standard Google's advertising business. Still, the hard work I can be advantageous in many ways.

Because Google's online ads everywhere, whatever makes people easy to use Web profitable company. And the system could lead to new applications. A few weeks ago, the company said he would use voice recognition to create text-captions YouTube videos in English, which could then be translated into 50 other languages.

"This technology can make the language barricade is gone," said Franz Och who is chief scientist at Google who heads the company's machine translation team. "It could permit anyone to communicate with others."

Mr. Och a German researcher who previously worked at the University of Southern California, said he was initially reluctant to join Google, fearing it would treat translation as a side project. Be an founder Larry Page that Google only, call and convince Mr. Och.

"Basically he said that this is something that is very important to Google," Mr. Och just now. Mr. Och entered in 2004 and immediately took a promise from Mr Page into the trial.

While many translation systems like Google's use up to have billions of words of text to build a language mapping, Google uses much more: a few hundred billion words in English. "Mapping to get better and continue to improve if more text is processed," said Mr. Och.

Efforts ultimately benefit. A year later, Google won a competition organized by the government to test a variety of complex translation system.

Google has used a similar approach to the massive computing power, heaps of data and statistics, to solve other complex problems. For example, in 2007, the company began offering 800-GOOG-411 directory assistance services that are free that translates spoken words. This service allows Google to collect the various sounds of millions of people in order to be better in terms of recognizing spoken English.

A year later, Google launched a search system based on sound as well as made many years other companies. And at the end of last year, Google introduced a service called Goggles that analyzes cell phone photos, matching them with a database of more than a billion online images, including photographs taken different paths to service Street viewnya.

Mr. Och acknowledged that Google's translation system still needed improvement, but he said the system was rapidly getting better. "Quality improvement curve is still high," he added.

No comments:

Post a Comment