The European Patent Office (EPO) officially announced today that they will collaborate with Google Translate in the area of online patent translation.
This announcement does not come as a surprise. At the Translingual Europe 2010 conference that was held last June in Berlin (read the blog post I wrote at the conference), a panel from the European Commission talked about how the some of the EU organizations, such as the EPO, are using MT. Ms. Patricia Biani from the EPO stated that they are abandoning the rule-based systems and focusing their efforts on SMT. Ms. Biani mentioned that they were trying Language Weaver’s MT but I guess that Google gave a more compelling sales pitch.
Machine translation is not new to the EPO. Free online translation has been available on the EPO’s Espacenet website since 2007. The translation service used Worldlingo, which used a Systran rule-based (RbMT) engine. The system will now evidently use Google to generate the translations.
There are upsides for both the EPO and Google:
- Google will get their hands on a ton of high quality bilingual corpus. All patents which are submitted to the EPO must have the Abstract submitted in the EPO’s three official languages: English, German and French. Google will get their hands on these corpuses and use them to train their translation engines.
- The EPO will enjoy better quality machine translation once Google does the training. The EU spends fortunes of money on translations and this will help them reduce costs.
It is unclear whether Google will get its hands on corpus in the other languages. National filing of patents in the EU member states requires translation of the entire patent into the local language. However, those documents are not always submitted to the EPO. And most patents do not get translated into all of the EU member languages.
Furthermore, many inventors file patents with the WIPO under the Patent Cooperation Treaty (PCT). The EPO and WIPO have an agreement under the PCT but it is unclear whether the EPO has access to the WIPO content database.
The EPO-Google deal comes at a time when the EU member states are bickering over translation requirements (read about that here). Many EU technocrats argue that the cost of filing a patent in Europe is 10 times the cost of filing in the USA due to translation costs.
Can Google machine translation replace the human translation process? Not likely. A patent is a carefully worded legal document- the wording of the claims can make or break an IP infringement case. Will Google replace the need for national patent filing in the language of each EU member state? That seems more of a political issue than a technical one. But this is good news for inventors and patent lawyers as translated search of patents is sure to improve.