15 May 2017 - No comments »
Since 2013, the EPO has offered Patent Translate, a free to use translation tool that allows anyone interested in patents to gain access to a world of patent documentation in other languages. Recently the service has undergone a change that has improved this service even further, thanks to a new technology called Neural Machine Translation (NMT).
Traditionally, we’ve offered instant translation of patent documents for 28 official languages. Now, using an artificial neural network developed by Google, NMT has been applied to patent data in eight of these languages, in addition to English (Chinese, French, German, Japanese, Korean, Portuguese, Spanish and Turkish). Essentially, weighted significance is applied to each word, based on context and meaning and the system is able to continually learn. In terms of both accuracy and fluency, the new neural machine translation solution is producing significant improvements in the quality of translated texts.
This remarkable progress has been due in part to the long standing co-operation that has been fostered actively between the EPO and Google. In a recent speech, one of Google’s machine translation experts stated plainly that working together with the EPO had been a key step in achieving the quality that Google can offer with the machine translations. One aspect of this success is the sheer amount of data that the EPO has been able to provide. A prerequisite for good machine translation is the existence of high-quality human translations between the two languages concerned. The computer can analyse these human translations and learn from them but to do so, the computer needs a lot of translations. In fact, the EPO has set a threshold of several tens of thousands human translations in a language “corpus” before it considers offering the language in Patent Translate.
It is impossible to overstate the importance of Patent Translate in a patent system which relies on knowing whether or not similar inventions already exist elsewhere in the world. In fact, one of the very purposes of the patent system is to disclose new technologies to others. For over a century, these disclosures have only been accessible to people who could understand the language they were written in. Thanks to Patent Translate, the EPO can offer the content of patent documents from all over the world to users everywhere. In choosing Patent Translate, they are assured not only of quality, but also of a secure service protected from eavesdroppers. Those using the service have the peace of mind that their online behaviour will not be analysed and their rights to data protection upheld.
Patent translation not only supports innovation by providing access to technical information, but it also represents a massive potential leap in productivity and efficiency. Back in 2013, one study we carried out led us to estimate that it would take 16,000 man-years to translate the Chinese patent documentation available at the time into English. Patent Translate provides all that documentation instantly – and the additional documents published since then – in English, French and German. They also have a level of quality that makes it easy for readers to understand the nature of the invention.
Our usage statistics are also showing that the investments we make in Patent Translate can have an effect not just in Europe but on a global scale. A large proportion of the 15 000 translation requests we receive on average every day come from India, Japan, Russia and the United States, in addition to those that come from the EPO’s member states. It is also those same EPO Member States that have been instrumental in supporting these developments by providing patent documentation.
Neural Machine Translation takes automated translations to a new level. In the future, more and more languages will be using NMT technology so that we can continue to bring cutting edge translation services for patents to all those who need them. You can try the service for yourself here. I am convinced that you will be as impressed as I am with the quality of the translations.