From the EPO website (link):

"30 November 2010

The EPO and Google have today signed a Memorandum of Understanding to improve access to patent translations in multiple languages.

Under the agreement, the EPO will use Google's machine translation technology to translate patents into the languages of the 38 countries that it serves. In return, it will provide Google with access to its translated patents, enabling Google to optimise its machine translation technology. Google technology will be used to translate patents originating in Europe as well as patents originating in other regions of the world and enjoying protection in Europe."


"For Google, the collaboration offers a major opportunity to improve its translation service. The EPO will offer access to around 1.5 million documents, and each year this number grows by more than 50 000 new patent grants. The partnership also covers Asian languages. Facilitating access to the rapidly increasing volumes of Japanese, Chinese and Korean technological information is one of the biggest challenges facing the global patent system."


Personally, I'm no fan of the Google patent search engine since the list of things it can not do far outweigh the little bit of functionality it provides.  Adding the 1.5 million EP patents will make it more useful though and increase the need to use it as a double check (especially since our clients will be using it).  If it improves Google's translation abilities with regard to patent and scientific documents, then something very good does come out of it for us.

And who knows, maybe this means that Google will someday actually improve their patent search engine...


  1. Dominic,

    Thanks for posting this news to PIUG. My colleague Chris posted about this over on the Intellogist Blog with an eye towards what this could mean for the future of a pan-European patent - It looks like this may be a stepping stone/trial program to making the translation process less of a burden on application filers who do not practice one of the three official languages (English, German, and French). If you're interested, our post and Chris's analysis can be found here:

    I could be wrong, but to me it does not necessarily appear certain that Google Patents will incorporate the 1.5 million document EPO file. I think the passage sounds like that Google is using this data for the refinement of its machine translation software only. From that perspective I don’t think we need to worry about Google Patents' flaws any more than we already do. It's possible that Google will expand beyond the US patents they currently host, but I'm not sure if that product is really serving them very well as any kind of moneymaker.


    1. My take on this is to agree with Kristin.  My interpretation of the press release is that this exercise is part of the ongoing attempts to find a resolution to the thorny problem of the working language(s) for the EU patent (ex-Community Patent).  The fact they are talking about "29 languages" says to me that the work is focussing on translating EP-B specifications into the national languages of the designated states.  This is currently done at the expense of the applicant, and is a major cause of grievance that the London Agreement was supposed to alleviate.  So this work is an attempt to train the translation engine to work from Eng/Fre/Ger into other languages such as Finnish, Danish, Spanish, Portuguese, Italian etc. If the system is perfected, it may be that these MT will be acceptable to ensure the the patent enters into force in these states, rather than using human translations.

      I take this view because the statement that "[Google] will work on translating about 1.5 million patent documents in 29 languages on behalf of the Munich-based EPO" [sic - courtesy of the Bloomberg release] doesn't make sense if it is supposed to mean that the source documents are ones published by the EPO, since the EPO only has 3 official languages.  Neither can it mean that the collections from the national patent offices are the source, and they are being translated into the EPO's 3, ; it would not be up to the EPO to sanction this work, it would have to be approved by national offices.  So I conclude that it is focussing on intra-Europe translations of granted EPs. The sentence in the EPO release which says "the EPO will use Google's machine translation technology to translate patents into the languages of the 38 countries that it serves" makes much more sense.

      With regard to making them searchable, I would interpret the statement in the EPO's own release to the effect that "The EPO will offer access to around 1.5 million documents, and each year this number grows by more than 50 000 new patent grants" actually means that they will offer 1.5m texts to Google to play with, not necessarily that the resultant texts will make their way into any new search file.


  2. Accepting the risk of adding a less informed comment (I have not yet read the EPO press release), I would like to add my two cents.

    The EPO application documents have value and possible future consequences for all its member states. Nevertheless, they are available in just one of the official languages. After the London agreement, it is not necessary anymore to translate the entire granted document into all languages of the member states. As a consequence, a lot of European patent (applications) are not available/searchable in full text in all languages of the member states.

    Of course we cannot exclude that this is a test of EPO to experiment with machine translations as a source for the official translations. On the other hand, it could also be an attempt of the EPO to make the full text of all European patent documents available/searchable in all languages of the member states.