GoogleTranslate, Linguee, and Summaries on the WIPO Website – A Brief Comparison of Three Excellent Resources for Patent Translators

 

Life can only be understood backwards; but it must be lived forwards.
Søren Kierkegaard

There was a time when the only thing that this poor patent translator had at his disposal was an expensive, non-interactive, inanimate object made from dead trees called dictionary. I still remember the times when I had to print documents for translation that were faxed to me from Japan among other places on curly thermal paper which often rendered Japanese characters illegible. The only thing that I could use to figure out the technical terms were overpriced dictionaries, which were generally obsolete by the time they were published. One (1!) such Japanese dictionary set me back 800 dollars in 1991.

I still have all of my dictionaries arranged in bookcases lining the walls in my office and in the hallway and sometime I still use them, but not very often. It is much easier and faster to use online resources mentioned in the title of this post and in contrast to paper dictionaries, online resources are free and they are generally updated very frequently.

Online resources available to patent translators include also search functions that can be accessed for free on the websites of the Japan Patent Office (JPO), European Patent Office (EPO), German Patent Office, French Patent Office, etc. A post that I wrote about these resources in 2010 is based on a chapter that I wrote for the Patent Translation Handbook published by the American Translators Association in 2007. Some of the information in that post and the Patent Translation Handbook is outdated by now, but I believe that most of it is still applicable and useful.

Like most translators, I prefer to use different online dictionaries and online resources, and I use different resources depending on which language I am translating.

Most translators by now probably have had plenty of experience with GoogleTranslate (GT). Regardless of which language I am translating, I usually go to GT for a first basic reference for a term that I plan to be using in my translations of patents, or technical articles and other documents.

However, translators need to be careful when they are using “free” machine pseudo-translation tools available online, or even paid tools, such as Adobe file conversion tool, which is what I use to convert long documents in PDF format to MS Word format, mostly to estimate the word count to provide a cost estimate.

Since there is no such thing as a free lunch, these free tools are not really free. Whenever we use a “free” tool, information about what we are doing online is generally the product that is being sold to other people, mostly for advertising purposes (or at least I hope so). Since translators don’t know who can have access to the information that they input online, they must be very careful when using pseudo-translation and file conversion tools. As described for example in the English version of the Japanese newspaper The Yomiuri Shimbun, confidential E-mails to a client and other documents that were translated by a careless lawyer through a free online service have been leaked onto the Internet and are visible to the public.

There is no need to worry about the text of published patent applications, because once they have been published on a website such as the JPO, EPO or WIPO website, they are in public domain. But texts of unpublished patent application may not be disclosed through a “free” online machine pseudo-translation service, or through a paid file conversion utility, because such an inadvertent disclosure could cause major problems for the translator.

Unless the document being translated is already in public domain, translators who use free machine pseudo-translation tools online need to carefully anonymize every document first by removing all names and any other identifying information from the document.

With the exception of music files on my i-Pad and i-Phone (where the operating system does not really give me any choice), I don’t store any of my files in the Cloud because I have no idea who will have access to my files. Even if the company offering Cloud storage promises complete confidentiality, they could be lying, or the files could be hacked. Of course, my computer could be hacked into too, but there are simple measures that I can take on my own to try to prevent such an eventuality when I am in control of the files. If I store my files in the Cloud, I am as helpless as passengers flying in a plane with malfunctioning engines as far as the confidentiality of the information in my files is concerned.

But let’s get back to comparing machine and human translations available on Google Translate, Linguee and the WIPO website.

Google Translate

The use of GoogleTranslate offers several advantages. First of all, everything is very fast. I can use a single keyboard to type words that I want to bounce off GoogleTranslate in several languages, including German, French, or Russian, and GoogleTranslate will generally figure out the correct accents and fix my spelling mistakes. With Japanese and Czech, I generally load a Japanese or Czech keyboard into the computer’s memory first. Given how complicated typing in Japanese and Czech can be, it is faster to use a special keyboard for these languages.

The advantage of GoogleTranslate is that while it provides machine translation, or machine pseudo-translation, which is probably a better term to use because what a machine offers to a human reader is not really a translation, unlike previous machine translation systems that were based on an unmapped minefield consisting of grammatical and syntactical rules combined with a dictionary, GoogleTranslate instead attempts to identify a previous translation that was done by a human and that closely matches the source text for translation.

I found this approach to be incredibly effective on many occasions, for example when I was translating Japanese laws and statues, because most of the text is usually completely identical to the previous version of the law and all I have to do then is to copy and check identical portions and add the new text.

But even this approach can often result in completely and hilariously nonsensical mistranslation when the match identified by the software is a mismatch, which the machine does not know. Since machines obviously don’t know anything as they just do what they were programmed to do, they can be trusted about as much as presidential candidates, by which I mean that you have to assume that what they are saying might be true, but there is also a good chance that everything they say is a lie.

In a larger portion of machine-translated texts, there will always be mistranslations, which may be difficult to detect even if you know the original language and impossible to detect if you only know the target language.

When I threw the German term “Verstellweg” that I was not sure about in a patent translation that I was working on at GoogleTranslate, I got back: “adjustment”, “adjusting”, “adjustment path”, “displacement” and “displacement path”, along with definitions of the words in English and German, which I did not really need. Most of these terms would kind of work in the patent I was translating, but “adjustment range”, which was in my opinion the best translation for this particular design, was not listed.

Linguee

Unlike GoogleTranslate, Linguee is an online dictionary search engine, not a machine translation engine, which means that instead of translating (or pseudo-translating) whole chunks of text, it displays existing translations of words that human translators may be looking for in different contexts. Incidentally, I find it interesting that the concept of Linguee was developed by a former Google Employee.

Just like GoogleTranslate, Linguee can be used for many languages, including Japanese, Russian, Czech, German, French, etc., and it also guesses, usually correctly, the proper spelling in other languages when I use US English keyboard for example to write text in Russian or Japanese.

This is very convenient, but I find it slightly creepy. If things continue like this, nobody will be able to remember how to spell anything in a few years. Why bother when the machine does it for us? Especially with languages using a very complicated writing system like Japanese or Chinese, young people probably already forgot how to write properly characters since their computer or smart phone remembers all those complicated characters for them. Human brain remembers only what it needs to remember, that is simply how it works. It is almost as if our brain had its own brain that does not really listen to us that much.

When I threw “Verstellweg” at Linguee, I got back 34 examples of sentences using this term in different contexts. Each example is provided as a full sentence in both languages with the queried term highlighted, which makes it easy to quickly scan all the examples to find the best possible match.

The great advantage of Linguee is that unlike GoogleTranslate, it provides a great deal of context which would not fit into the design of a machine translation engine. Just like in life, context is everything also in translation, and thanks to the simple and effective design of Linguee, searching for alternatives is usually quite fast.

I really, really like Linguee.

Summaries on The WIPO (World Intellectual Property Office) Website

I have been using this website for many years, longer than GoogleTranslate and much longer than Linguee. While the European Patent Office (EPO) website can be searched only in English and the Japanese Patent Office (JPO) can be searched only in Japanese or in English, millions of patent publications contained in the WIPO database can be searched in many languages, including English, German, Japanese, Russian, etc. A search for a term in a foreign language will usually identify English abstracts of many patent documents containing an English translation of the term in German, French, etc. Because complete summaries are displayed, even more life-saving context is contained in the summary.

However, sometime there is no English summary and only a summary in the original language will be displayed, presumably because a translation is not available yet.

The English summaries are usually quite good, but not always. This is probably due to the fact that WIPO sends these summaries for translation to various translation agencies and while good agencies will work with really good translators who do good work, other agencies may use really lousy translators who are generally much cheaper.

Once in a while I find that the summary that is provided in English along with a summary in a foreign language sort of explains the main principle of the patent publication, but complicated terms that I am looking for are completely missing in English because a dishonest translator simply did not bother to translate the complicated terms. On the one hand, I think that is unforgivable for translators to behave this way, but then again, given how much (or rather how little) they may be paid by some translation agencies, and how little time they may have to translate a large number of abstracts in order to pay their bills, I would blame more the translation agency than the translators.

When I threw my test term “Verstellweg” at the WIPO website, I got back 608 hits. While this sounds like even more context, the problem is that trying to identify correct English translations takes a long time because only the word in German is highlighted, and sometime when I click on the German text, the summary is not available in English.

I find myself using the WIPO website much less now, mostly because it takes such a long time to find what I am looking for among the hundreds of documents displayed. But the WIPO website is still a very good resource for me, and often I do find on it the answer to my question when my search was fruitless on GoogleTranslate or Linguee.

**********

Unlike two or three decades ago, translators now have many online resources that can be used to find correct terms. However, this does not mean that thanks to the availability of abundant resources, just about anybody who knows a foreign language can translate just about anything.

In order to translate and do it well, you still have to be a translator first.

I remember how Tom Petty was describing in a documentary about his music the primitive method that he was using to learn new songs when he was a teenager in the seventies in Florida. He said that he used to sit in a car listening to a song that he liked on the car cassette player while stopping it every few seconds to write down the lyrics and try out the guitar riffs on his guitar.

It is much easier for teenagers these days to learn how to play songs because everything can be simply downloaded from the Internet. But just because they can download what they need in a few seconds does not mean that this will turn them into talented musicians.

And just because there are many online resources available to translators these days, that does not mean that just about anyone can become a translator.

Whether you want to be a musician or a translator, you have to have the music in you first, because otherwise it is probably not going to work.