Patents and Open Source: Having Your Cake and Eating It, Too

Patents and Open Source: Having Your Cake and Eating It, Too

Contrary to popular belief, protecting your business through patents is not necessarily at odds with an open-source approach.

The two can be combined with successful results. One dynamic start-up working in machine translation software filed a patent in its early days—but the inventor still refined the technology via an open source platform, helping to turn the technology into an industry standard.
Patents have many uses. They are more than a barrier to keep rivals out. They can be the basis for fruitful collaboration or vital commercial intelligence.  They can also help you build brand value. Patents are an asset but also a tool. For small companies and start-ups, patents may convince wary investors in the cautious money markets. The IP in any innovative project, and how it is managed, can be the dealmaker: in the early stages of a new venture, it might be the only thing that can secure finance.


Take the example of German computer scientist Philipp Koehn, who together with his professors at the University of Southern California, developed and patented a new model for statistical machine translation.


Translation by computers had been around for a long time but, until Koehn came along, machine translation had not come close to reaching its full potential. Koehn and his team came up with a phrase-based model (translating words in their context, rather than one word at a time), which was a cut above the old sluggish word-based systems. Their new technique found the best statistical match for words based on how often they appeared in pairs of existing translations. The results were astounding. Today most of the big names in Internet translation have integrated this technology. For their invention, Koehn’s team was nominated for the European Inventor Award 2013.


By patenting their technology, Koehn and his team could secure crucial financing (from both venture capitalists and public research programmes), and after seven years their company, Language Weaver, was bought by SDL for $42.5 million. But Koehn later opted for an open source approach, convinced that it could lead to broader research, development, and usage. At the University of Edinburgh, he started an open-source platform called Moses, which he still runs today. A community of enthusiastic researchers around the world can directly access and contribute to the software, effectively crowd-sourcing improvements.


The economic benefits are two-fold: Firstly, many companies have already integrated the free open source technology into their organisations, creating a big user- and client-base. Secondly, a multitude of new software companies have been founded, expected to turn phrase-based statistical machine translation and related services into a several billion-dollar market in the coming years.


Koehn understood that there is a huge world market for ‘imperfect’ translations that allow people who need a quick translation to get the gist of what is being said. Today Moses is used by big companies for document and website localisation, so customers all over the world can read their instruction manuals or properly install a computer programme.

Patents on software?

This example shows that the traditional proprietary patent model is not incompatible with open source. But it also debunks another myth, mainly about so-called “software patents.” It is true that the lines of code in a programme are protected only by copyright (like the lines of text in a story), but not by patents. However, a great many inventions we know today, like GPS, wi-fi, Bluetooth and mobile phones all rely on software. If they are new and inventive or solve a technical problem (“How can I use satellites to guide me to my destination?”) they can be patented.


Such patenting is possible when the product or process of the invention offers more than mere lines of software. The resulting patent protects the invention regardless of the particular lines of code or computing language used in the underlying software. Koehn’s phrase-based model of statistical machine-based translation is one of these “computer-implemented inventions.”


Of course, the value of non-patentable inventions cannot be denied. Much of the service industry sector, and the creative industries, have little use for patents, despite innovating to a very high degree. After all, only a minority of businesses are innovating in sectors where patent ownership is relevant. Even so, there are many other ways in which the patent system can be used to create commercial advantages.


Oswald Schröder


How Do You Find Your First Customers?

How Do You Find Your First Customers?

As any business owner will tell you, finding those first ten or twenty customers is always the hardest. For one, yo