According to an EPO announcement on June 5, 2013, “The EPO and the Japan Patent Office (JPO) today announced the launch of the Japanese-English component of the EPO’s automatic translation service Patent Translate. This means more than a million Japanese patent documents available via the EPO’s global patent database Espacenet can now be instantly translated into English free-of-charge at the click of a mouse. This major step offers access to Japanese patents in the full-text version, while Japanese inventors can read European patents in their own language.” …. the JPO has taken a wide variety of initiatives such as enabling the use of machine translation in order to break through the foreign-language barrier and support users when they file international applications and enter overseas markets”.
Every time I read an important announcement like this, my heart beats a little bit faster, a smile appears on my face and I feel like I should open a bottle of wine to celebrate my good fortune as I happen to know that the easier it is for patent lawyers to access machine-translated patents, the more work there is likely to be available for humans who translate patents such as this mad patent translator.
Google Translate is an excellent MT product. I use it every day, as do thousands of other translators. But although Google Translate will give you some idea about a patent in a foreign language, sometime a pretty good idea, it cannot be really used to translate patents, especially from a language as complicated as Japanese.
During the last few years I replaced a stack of technical dictionaries that used to adorn my desk with Google Translate and other MT engines. I usually print out the machine translation and during the initial translation stage, I am constantly looking at the printout while I am trying to establish the terminology that I will be using in my translation. After a few hundred words, I usually glance at the machine translation only occasionally, although I frequently type terms in Japanese, or German or French into Google Translate, sometime Microsoft Translator or another MT tool to help me with a word that I am trying to figure out.
In this respect, machine translation programs have made life for translators like me much easier. But while machine translation is a tremendous help for translators, it can only provide a limited utility for people who know only one language. If you know both languages, machine translation will save you a lot of time, and you can easily spot places in the text where MT made mistakes and see the mistranslations. But if you only know one language, the barrier of a foreign language will be for the most part only replaced by a maze in your own language.
In Greek mythology, an enormously complicated structure called Labyrinth was designed by the architect Daedalus in order to imprison in it a man-eating monster called Minotaur, half man and half bull who was able to enjoy dining on young men intent on killing him in the Labyrinth for quite a while until one young man called Theseus decided to kill Minataur with the help of Adriadne, daughter of King Aegus of Athens. Nobody was able to get out of the Labyrinth, which made it easy for Minotaur to continue enjoying his dining experience until cunning Adriadne gave Theseus a magic ball of thread to tie to one end and follow it on his way through the Labyrinth.
What does a ball of magic thread from Greek mythology has to do with machine translation, you may ask? Why, everything! Designers of machine translation programs keep trying to “break the language barrier” by studying grammar and coming up with new algorithms, or by replacing the grammar-based approach with a huge database of existing translations in the case of Google Translate.
But the problem is, MT cannot find the way forward and then the way back again in the huge labyrinth of human languages without the magic thread that all humans naturally have and use all the time when they speak and think … the magic thread that humans call “meaning”.
Humans can naturally see how this magic red thread called meaning is woven into the words, spoken or written on a page. But because unlike humans, machines don’t understand the concept of meaning and never will, machine translation programs so far have only been able to make only limited incursions into the Labyrinth of human speech and human mind. Sometime they may keep going further for quite a while, but without the red magic thread that Adriadne gave to Theseus at one end of the maze to follow it to kill the beast and come back (she only gave him the ball of thread after he promised to marry her because like most women, she really wanted to get married), it is only a matter of time before even the best machine translation programs becomes stuck in the immense bowels of the Labyrinth.
You may still remember from Greek mythology that King Minos then got so mad at the architect Daedalus for being unable to create a labyrinth that would protect his pet man-eating Minataur that he imprisoned Daedalus and his son Icarus. Both of them in the end got out of the prison by flying on artificial wings made of wax … but as Icarus got too close to the sun, the wings melted and Icarus fell into the sea and died.
I am not quite sure what the moral of the story about Icarus would be with respect to machine translation. Oh, I think I’ve got it: the moral of the story is that people who talk about yet another software program that “breaks through the language barrier” need to make sure first that their wings are not made out of wax.