Welcome to the Machine
A bleg of sorts concerning machine translation:
The European Patent Office (EPO) recently announced that it has reached an agreement with Google to machine-translate patents. As a semi-professional translator (not of patents thank God), I don’t know whether to laugh or cry: machine translation is just awful, useless in every regard (if patents themselves aren’t). Every professional translator knows this, yet somehow Google has convinced the world – and the world’s governments – that its machine translation technology has value other than as a simple reference tool (and a third-rate reference tool at that).
I’d like to write a full-length (10,000-words-plus) article about the EPO’s decision and try to get it published in a major magazine. Nevertheless, I’m sorely ignorant in more than a few relevant areas, namely international patent law and machine translation programming. I’d like to either consult with experts in these fields during the process of writing the article or – preferably – co-write the article with these experts and have all of our names appear in the by-line.
I figured I’d throw it out here on the sidebar at the League since we tend to collect the types of people who would have this kind of expert knowledge or who would at least know other people with this kind of expert knowledge, and I’m already kind of speculating on which of the commenters might be interested… so please contact me at christopher.carr1984@gmail.com if you’re interested in the project.
And of course, the replacement of human beings with machines is always interesting discussion fodder, so let the techo-accolades of techno-acolytes and the lamentations of Luddites commence in the form of comments.
Oh, I was expecting a piece on Pink Floyd.Report
I work IT for an IP law firm and right now most everyone is just using Google Translate for foreign language documents anyways (or worse, they run the document through OCR and THEN through Google Translate!) My limited non-IP lawyer understanding is that generally, if you’re referring to a foreign patent document, then you’re really just looking for the gist of it to see if it bears any relation to whatever you’re working on for purposes of prior art. Nuance is rarely needed (and if it is then they have translators they ship the documents off to). So this doesn’t really surprise me.Report
Also worth mentioning that they’re not going to be using off-the-shelf Google Translate — it’ll be specially trained for this content. Assuming that patents include a lot of boilerplate and customary language (and little room for personal style), it might even produce halfway-decent results.Report
Christopher, I’m not an expert, but I think international patent law has become pretty standardized, due to the WTO TRIPS (Trade Related aspects of Intellectual Property Rights) agreement. In the absence of finding an expert quickly, you might find value in the WTO’s TRIPS page.Report
Machine translation might work reasonably well in this context if its lexicon was compiled exclusively from legal or quasi-official documents, say UN translations. James has already referred to TRIPS page. Such a paradigm would exclude vernacular usages.
The underlying problem is fairly obvious: the lexicon of Law French is already resident in English and American law but the modern usages in both French and English are profoundly different. The vocabulary of law is horribly intransigent in every country, refusing to evolve in the face of modernity.Report
I would be happy to collaborate with you on thisReport