Search:  

 Blog     e-Newsletter       Resource Library      Directories      Webinars     Apps
AFCEA logo
 

Talk Is Cheap

September 2011
By Max Cacas, SIGNAL Magazine
E-mail E-mail the Author

 

Google Translate has an app for the Android smartphone.

Europe turns to a commercial search engine to untangle Patent Office tower of babel through no-cost partnership.

The agency that administers patents among the nations of the European Union is about to receive some high-technology help from across the Internet. And, it will do so without a single euro changing hands. Late last year, the European Patent Office agreed on a memorandum of understanding with Google to use the California-based company’s machine translation technology to help process patents into 28 European languages, as well as into Chinese, Japanese, Korean and Russian. Members of the European Union ratified and signed the formal agreement this past spring.

Richard Flammer, head of Patent Information with the European Patent Office (EPO) in Vienna, Austria, says it is an effort to deal with what he describes as an information tsunami.

“In our present database, we currently have stored 70 million documents,” he explains. The EPO’s public website through which that database is accessed experiences 30,000 visits per day, primarily from patent attorneys researching the intellectual property status of new products versus old. In addition, official EPO patent examiners use the site to determine whether or not to award patents to new products. The site logs an average of 100,000 user sessions per day; all of those users download an average total of 100 gigabytes of data, or approximately 11 million pages daily.

Then there is the cost involved in translating the results of a patent search from the huge multilanguage database. Offering one scenario, Flammer suggests that a patent attorney researching a client’s new product might discover documents on as many as 150 other similar products. Flammer says some returned documents could be in any one of the many languages from EPO member nations, and, in some cases, English may not be one of those languages.

At current rates, it costs approximately €80 per page to translate patent information from one language to another at an officially accepted European Union (EU) translation agency. A spokesman for the EPO says the translation for one patent alone could climb to as much as €1,500, or approximately $2,100.

Flammer says the goal of the Patent Search application now being developed by Google is to reduce both the cost and the time it takes to conduct the average patent search. To obtain a translation, Flammer explains, “You make a query in Espacenet [the EPO’s patent search website]; you find a document; and you press ‘translate.’ In no time at all, you find the conversion in whatever language you want, in a quality that you can understand.”

Flammer explains that the move to a fully online capability for the translation of patent information from so many different languages is part of a generational transition from paper, to CD-ROM, and then to the Internet.

Flammer relates that several years ago, users at first rejoiced at the free availability of patent information on a Web-based database system, but then they realized they had what he describes as an information tsunami on their hands.

“We had millions of patent documents in all sorts of languages,” he says, “and no one could understand the content of the information.” Flammer reports that the development of Google Translate’s Patent Search is less about “patent information than it is about patent intelligence. It’s about finding the relevant information to make relevant business decisions.”

The EPO currently includes 38 member nations, speaking a total of 28 different languages. Three of those languages—English, French and German—are considered the three main working languages of the EPO, and all patent applications must be submitted for official consideration and approval in one of those three languages.

“From Finnish to Greek, from Spanish to English, from German to Italian,” Flammer reports, the goal of the new Google-EPO partnership is to ease the multinational patent process.

“We would like to have all those languages translated into English by the end of this year and all those languages into the three working languages—English, French and German—by 2014,” he says. Because 70 to 80 percent of patent requests are submitted in English, the first phase of the Patent Search will yield English translations, according to Flammer.

The eventual goal is to expand the scope of Patent Search to nations participating in IP5, an international forum representing the world’s five largest patent-granting offices. Along with the EPO, they are the Japan Patent Office; the (South) Korean Intellectual Property Office; the State Intellectual Property Office of China; and the U.S. Patent and Trademark Office. These offices handle 90 percent of the world’s patent applications.

If successful, the capabilities of Patent Search will be extended to Chinese, Japanese, Korean and Russian, under the terms of the Google/EPO agreement. Just as the EPO has done, the IP5 has placed a high priority on successfully developing machine translation of patent information for searches and the patent granting process.

Flammer acknowledges that Google’s Patent Search is not going to put human translators out of business, at least not in the patent industry. “It’s very clear that the quality of a machine-translated text may never reach the quality of human translation,” he explains. Rather, it provides a tool for patent attorneys and examiners to obtain a first-pass translation of a patent search that can be refined more readily into a legally binding intellectual property document by a trained human translator.

The nonexclusive agreement signed in March means that in exchange for free machine translation services, the EPO will make available to Google for purposes of developing the Patent Search application, its corpus—or complete database—of translated patent documents. The existing documents will be used by Google to further refine and optimize the abilities of its server-based language translation algorithms.

Jeff Chin, product manager for Google Translate, characterizes the agreement with the EPO as a win-win for both organizations. The arrangement is another example of a practical application of cloud computing, with users accessing Google Translate’s Patent Search via the Internet on the vast, global network of servers providing everything from its search engine, to YouTube videos, email and maps.

“We provide access to our translation system, our translation API [application programming interface]. They can make that patent information more accessible,” Chin says. “One way that we learn is from previously translated documents. By incorporating this parallel—EPO—corpus, we will improve the translation quality of our system.”

 

The Patent Search application now being developed for the EPO is expected to help patent attorneys and examiners save time and money conducting patent searches in a multilingual patent database.

Since 2006, Google’s Patent Search has been providing a separate, but related, search and indexing service for U.S. patents and applications, allowing anyone including patent lawyers and the general public to read patents online or download them as PDF files.

Officials hope that Google’s Patent Search application will make possible another long-range goal: development of the unitary patent or a trans-national intellectual property protection system that works across the member states of the European Union.

Patent Search would become the underlying technology for a system that would allow for quicker research and development of new technologies by providing a one-stop shop for researching existing patents across multiple languages. Such a shop would reduce the time and expense needed for ensuring that new inventions and innovations in one country do not infringe on the property rights of others in neighboring nations.

Google Translate’s performance in the patent scenario promises many efficiencies, but in rural or battlefield environments, the application is hindered by both connectivity issues and the amount of data available to verify English translations to certain languages. Chin explains that Google Translate currently has few, if any, of the parallel documents containing verified English translations from languages such as Pashtun, a regional dialect native to those who live in southeastern Afghanistan and northwestern Pakistan.

Leaving open the possibility of integrating such tongues into Google Translate’s database, Chin adds, “If there’s an entity or an organization that has a large amount of parallel data, we’d be certainly willing to work with them.”

Mindful that language translation often takes place away from the comfortable confines of a desktop computer, Chin says that Google Translate does have a mobile application that runs on handheld devices using the Google-developed Android operating system.

“We do have the Android Translate app, available for free in the Android Market,” he states, and “it supports all the languages available at translate.google.com. In addition, we’ve integrated our voice recognition, and text-to-speech, so that you can speak into the phone.” The application would transcribe the user’s voice, perform the translation and then speak the translation out loud. The application has been refined to allow limited language translation for conversations between two people who speak English and Spanish.

WEB RESOURCES
European Patent Office: www.epo.org/index.html
Google Translate: http://translate.google.com/
EPO Espacenet Patent Search Engine: http://worldwide.espacenet.com
IP5 International Intellectual Property Forum: www.fiveipoffices.org/index.html