IP TRANSLATOR, or why MY BRANE HERTZ

If the words IP TRANSLATOR don’t immediately ring a bell, don’t worry. Probably it just means that you are not a EUropean trademark lawyer. (Neither am I, but I play one on TV.) IP TRANSLATOR, or more formally Chartered Institute of Patent Attorneys v Registrar of Trade Marks (C-307/10) is an important trademark case decided by the EU Court of Justice earlier this week. To help you understand what I’m ranting about, let’s start with some background. (Oh, and it’s IP as in Intellectual Property, not Internet Protocol.)

1. The two essential components of the registration of a trade mark are (a) the sign and (b) the goods and services which that sign is to designate. Each of those components makes it possible to define the precise subject-matter of the protection conferred by the registered trade mark on its proprietor.
(from Advocate General Bot's Opinion or draft judgement)

What this means in practice is that the same word(s) can be a registered trademark for several different owners at the same time, as long as the products are different enough. High-reputation marks are a bit different in that they receive a broader scope of protection because anyone else trying to register, say, ROLLS ROYCE for anything whatsoever has certainly chosen the mark to try to gain unfair advantage of the earlier brand’s reputation. Even if many of the trademarks you can recall might be high-reputation marks (and certainly all the most valuable brands are, which is kind of the whole point), the vast majority of trademarks in the registries are not high-reputation marks, and thus must coexist with similar marks of others’ if necessary. In the EU trademark registry at OHIM (Office for Harmonization on the Internal Market (Trade Marks and Designs)), at the top of the list (apart from single-character renditions of picture marks) there are right now 90 different entries (granted/pending/rejected applications) for FUSION, also 90 for ECLIPSE, and 89 for EVOLUTION.

In order for the registrars (and everyone else) to keep track of the different products for which marks are registered, they are classified into 45 different classes according to the Nice Classification, 34 classes for tangible goods and 11 (up by three quite recently) for services. The classes are quite uneven in their scope, ranging from

Class 9. Scientific, nautical, surveying, photographic, cinematographic, optical, weighing, measuring, signalling, checking (supervision), life-saving and teaching apparatus and instruments; apparatus and instruments for conducting, switching, transforming, accumulating, regulating or controlling electricity; apparatus for recording, transmission or reproduction of sound or images; magnetic data carriers, recording discs; compact discs, DVDs and other digital recording media; mechanisms for coin-operated apparatus; cash registers, calculating machines, data processing equipment, computers; computer software; fire-extinguishing apparatus.

to

Class 23. Yarns and threads, for textile use.

The classification also contains an Alphabetical list of over 8000 more specific product descriptions divided into the individual classes (637 in class 9; 22 in class 23). Overall, in my opinion, the Nice Classification stands out as a relic of a bygone era designed to fully leverage the state-of-the-art computing power of index card technology. (Using words like ‘gutta-percha’ in the class headings doesn’t exactly help.) Of course, fifty-odd years ago the registries were also much more manageable in size.

In EU (at OHIM and in most of the Member States), it has been possible to register a trademark not only for a detailed list of specific goods and services (obviously covering only those) but also for all goods and services in an entire class by using the class heading as a product description. This is different from eg the US, where only specific products must be used, and trademark protection normally only extends to those products for which the mark is actually used in trade, even if the original registration also covers additional products.

Another thing to remember is that a trademark must be able to distinguish the owner’s product from those of others’. Because of this, a trademark cannot just be descriptive and quite simply name the product by its common name. Trademark degeneration, or a trademark becoming a generic name for a certain type of product, can mean that the trademark is no longer protected. (This has happened for example for ASPIRIN in the US, whereas in Europe Bayer’s registration is still valid.)

In the IP TRANSLATOR case, the CIPA applied for registration of IP TRANSLATOR as a UK trademark for ‘Education; providing of training; entertainment; sporting and cultural activities’ in class 41, which also happens to be the class heading for that class. The UK Registrar refused the application based on OHIM’s classification guidelines (Communication 4/03) on the grounds that IP TRANSLATOR is descriptive for translation services, which are also a part of class 41, even though they certainly do not naturally fall under any of the categories listed in the class heading in normal parlance. Of course there was an appeal, and this is where the ECJ was asked for a prejudicial opinion. And thus the ECJ opined:

[...]trade marks must be interpreted as meaning that it requires the goods and services for which the protection of the trade mark is sought to be identified by the applicant with sufficient clarity and precision to enable the competent authorities and economic operators, on that basis alone, to determine the extent of the protection conferred by the trade mark.

Directive 2008/95 [...] does not preclude the use of the general indications of the class headings [...] to identify the goods and services for which the protection of the trade mark is sought, provided that such identification is sufficiently clear and precise.

An applicant for a national trade mark who uses all the general indications of a particular class heading [...] to identify the goods or services for which the protection of the trade mark is sought must specify whether its application for registration is intended to cover all the goods or services included in the alphabetical list of that class or only some of those goods or services. If the application concerns only some of those goods or services, the applicant is required to specify which of the goods or services in that class are intended to be covered.

In response to this OHIM issued a new Communication 2/12 to supersede the earlier 4/03, requiring for future and pending applications one extra bit of information for each class of goods and services using complete class headings, namely whether it is intended to cover the entire class or not. This alone would only address the issue raised in last paragraph of the ruling, and ignore the first two altogether. Furthermore, the second paragraph in fine read together with first one quite simply means that not all class headings are created equal. The ones that are sufficiently clear and precise standing by themselves (like class 23) are fine, but others (like class 9) are not. This would of course bring registration practices more in sync between the EU and the US, which some might argue is a Good Thing. Actually this is even spelled out quite clearly earlier in the judgement:

54 In that connection, it must be observed that some of the general indications in the class headings of the Nice Classification are, in themselves, sufficiently clear and precise to allow the competent authorities to determine the scope of the protection conferred by the trade mark, while others are not such as to meet that requirement where they are too general and cover goods or services which are too variable to be compatible with the trade mark’s function as an indication of origin.

55 It is therefore for the competent authorities to make an assessment on a case-by-case basis, according to the goods or services for which the applicant seeks the protection conferred by a trade mark, in order to determine whether those indications meet the requirements of clarity and precision.

In response to this, OHIM is, for the time being, reviewing all incoming and pending applications using class headings on a case-by-case basis, while also trying to make up its mind as to which class headings are still okay, together with the local registrars. Even if registrations such as the class heading for class 9 are no longer allowed, the ones registered up to this point will of course still remain in force unchanged.

Of course my interpretation is far from being the only one around. From a trademark attorney’s job security perspective, this one by Niamh Hall as seen on IPKat would probably be the best one I have come across so far. Not only would class heading registrations be allowed for all classes, but they would only cover all products listed for that class in the Alphabetical list of the Nice Classification (and perhaps not even the latest one if the mark is older), and not just all the products that would eventually end up in that particular class, thus potentially excluding over 85% of the individual products listed in EuroClass, the product classification system used by OHIM and an increasing number of local registrars. From the perspective of legal certainty and predictability, this wouldn’t exactly be optimal.

(provide 'rant)

Lies, damned lies, and performance metrics

In the DESI IV e-discovery workshop at ICAIL last June, in one of the presentations there was a single slide that really caught my attention. It showed the performance of a large number of e-discovery systems on two axes representing the most common metrics used in information retrieval, computational linguistics, and what have you: precision and recall. (The slide was based an earlier publication which I unfortunately did not register at the time and have not been able to find since, pointers extremely welcome.)

What was so intriguing about that slide was that apart from a small number of outliers, almost all of the systems evaluated were at the bottom left quadrant, with both precision and recall (typically well) below 50%. Based on my expectations from my previous line of work, this was quite shocking, and had I known this earlier, I would certainly have worded a paper or two a bit differently as far as using e-discovery as an example is concerned. All the same, I suppose even at these figures e-discovery already outperforms all the alternatives, but there is certainly still considerable room for improvement.

What I was accustomed to was that a marketable product should deliver well above 90% in both columns or else the users would just stop using it. Whether they are 93% or 98% is not all that important, because that kind of variation is mostly just noise and depends on how well suited the test materials happen to be for that system. In particular, if you use a manually annotated test corpus as a gold standard, in a commercial scenario, you can really only use it once, because whatever is left between the actual performance and 100% are what we in the business call bugs, and they should be dealt with unless there is a good reason not to do so. And so our gold standard becomes tainted once these bugs are fixed. Which is of course no reason not to use the new figures for marketing purposes. (‘Press statements aren’t delivered under oath.’ - Jim Hacker, PM)

Performance is not just an e-discovery issue. It is raised in many other legal technology contexts as well. For example, the Swedish trade mark law start-up Markify prides itself with the ‘99% accuracy’ of its system, for example in this recent Arctic Startup profile. The actual study on which this claim is based is also available. The results are based on querying a set of 1000 actual cases of successful US trade mark oppositions and the question was whether the different services would return the correct mark (that of the opponent) when queried for the mark that is being opposed. Here are the results, and for added entertainment easier overall comparison I have also computed the F-scores for all of them:

System	Recall	Precision	F-score
Markify	99.7%	0.02%	0.04%
Thomson Reuters Compumark	45.5%	0.21%	0.4%
CT Corsearch	34.6%	0.43%	8.5%
USPTO	34.2%	0.31%	0.6%
Trademarkia	32.5%	0.75%	1.5%
CSC	31.8%	0.55%	1.1%

If recall is all you are after, improving on that 99% is really easy. Simply by returning the entire database for each query you can reach 100% just like that. At the same time precision naturally drops down to epsilon but so what. Of course this is not quite fair (there was no proper indication of the placing of the correct answer on the list of results), but still, just returning the desired answer is definitely not enough, at least when it is returned in a needle-in-a-haystack mode, where, even if the result is there, it is increasingly likely to be missed by the person reading the results the longer the list is. For what it’s worth, I tried to search for ‘äpyli’ (that’s Helsinki slang for ‘apple’) on Markify's system and quit after 10 pages of results at which time the trade mark of a well-known Cupertino-based fruit company had not yet shown up, and the results being shown at that point were already much further away. I suppose ‘the other [sic!] high quality paid trademark search services that can run $500 a word’ can still breathe easy.

What is Legal Technology?

In my opinion, AI & law desperately needs a makeover. One simple but effective way for the field to reinvent itself is rebranding, and the best label for this I can think of is Legal Technology (oikeusteknologia, rättsteknologi, ret(t)steknologi, õigustehnoloogia, R/rechtstechnologie, technologie juridique, tecnologia giuridica, юридическая технология &c).

As someone who has followed and worked in language technology for about two decades now, I see AI & law as being now in the same state as language technology was in the early 1990s. I have presented some lessons learned on how to approach real-world problems at the detail level in my robo-judge paper, so I won't go into them here. Instead, my proposal here looks into language technology as a field that has succesfully reinvented itself a couple of times already. Early on, it was only known as natural language processing (NLP) as a subfield of AI and a form of basic research rarely with any concrete application in mind. (A notable exception to this is machine translation, which also happens to be older than the term AI itself. More on that in a separate post as well as an article written jointly by me and Anniina Real Soon Now.) Then came computational linguistics, which was centered on using computational models and techniques as a tool for linguistic research. (This is where I think AI & law is now.) Of these, in particular corpus linguistics has become mainstream in virtually all subfields of linguistics, but other computational methods are now widely used outside computational linguistics proper as well. Through the 1990s computational linguistics also started to find its way into commercial applications in domains such as language checking, information retrieval, text-to-speech and vice versa, dialogue systems, and machine translation. As these real-world applications started to generate increasingly important research questions in their own right, language technology was born.

"Legal technology" as a term is not my invention. For example, in the US there has been a bicoastal biannual conference called LegalTech® since 2006. As far as I know, most of the technologies presented there are are not all that interesting from an AI & law perspective, with topics such as case management and billing platforms, synchronizing your BlackBerry with your Outlook and stuff like that, and whatever new cruft Westlaw and LexisNexis have come up with each year.

More to the point are for example the LawTechCamps arranged by Daniel Martin Katz (of Computational Legal Studies) and others in June 2011 and next week in Toronto and in the end of June in London. There is also a growing number of start-up companies in the field at least in the US, as listed just the other day on the eLawyering blog. Most of the start-ups listed seem to be working on applications having to do with contracts (possibly a sign of flock mentality from the venture capital side?). Contracts are also the target of the only out legal tech start-up I know of here in Finland, Sopima. With a large number of companies on the same domain fighting over the same market from somewhat different perspectives, it is clear that only some of these companies will be able to succeed (at least as far as the US companies are concerned, Europe is a different kettle of fish because of the different legal culture(s) and the prevalence of non-English languages). The best products have to address a real-world problem and solve it well and efficiently. Usability is another key success factor, and it still seems to be generally neglected in legal IT. Just because a certain design is a possible way to do something does not mean it is the best way (indeed it rarely is, though at least it usually is not quite this bad; required reading: Donald Norman's The Design of Everyday Things, MIT Press 1989). In particular, just replicating ancient practices from the age of pen and paper (and secretaries) and possibly adding some bells and whistles is a true and tried pattern, unfortunately. And the result is an application that takes a week-long course just to get started with it. All the same, the technically best solution does not necessarily win the game. In the end, it all boils down to the viability of the business model and the ability to make it into a reality. (Here's a convenient rule of thumb: marketing costs money, selling makes money. Close early, close often.)

So how can the AI & law community contribute to the impending legal technology boom? One approach is to take an existing, reasonably well-developed Good Old-Fashioned AI & law technology, and to find a real-life legal problem which it could possibly solve. (I'm afraid I can't come up with an example.) The other approach is to take an existing problem (= market need) in the legal community, a problem of the kind that should be solvable by computing, and looking around all over the place in computer science in search of that solution. (Here e-discovery is a prime example, though it does not travel well, and performance-wise it is quite disappointing by the language technology metrics I'm used to but at least it is still equally reliable yet faster and cheaper than people doing the same job.) Since language is in a key role in law, language technology is one obvious place to look at but it should definitely not be the only one for any legal tech company. I'm sure the next 20 years will be a lot more interesting (and profitable) for the field than the past 20.

Zombies and robots and cyclists, oh my!

Inspired by today's Decision by Rector Magnificus Thomas Wilhelmsson, I have made one of my own as well. Rather than keep struggling with writing a traditional monograph, I have decided to leverage my already published peer-reviewed work in the form of an article dissertation (like all the cool kids today).

That is, my thesis will consist of the obligatory introduction (which also includes a couple of things I could first publish as standalone articles but can't be bothered, namely looking at WEBSOM from a legal perspective and a business planresearch programme for cognitive legal theory), and the following articles, the first four of which are already published and the fifth is a written version (now with 100% more cognitive semantics!) of the paper I presented last summer at GikII:

• MOSONG: A Fuzzy Logic Model of Trade Mark Similarity
• From Spelling Checkers to Robot Judges? Some Implications of Normativity in Language Technology and AI & Law
• Dual-Process Cognition and Legal Reasoning
• Grundnorm as Conceptual Bootstrapping: An Ontogenetic Perspective
• Zombies and Legal Personhood
• Burdens of Proof, Probabilities, and Alberto Contador

This shouldn't take more than a couple of weekends per article plus some more for the introduction, so I might actually be able to defend my thesis in 2013 after all. I have a bunch of other article ideas in store as well, but they'll have to be put on hold for now. Stay tuned!

What is innovation?

I'm being profiled on WIN-novators later this spring. As an preview, here's my take on the question in the title of this post:

I suppose there are basically two possible aspects on innovation. The first is taking something preexisting and familiar and making it just a little better. The other alternative is a radical break with tradition and doing things differently altogether. In a sense this echoes the distinction between normal science and scientific revolutions introduced into the philosophy of science by Thomas Kuhn. But in reality (rather than philosophy) these aspects are complementary rather than dichotomous.

To take a concrete example: the shift from, say, a 32 nanometre to a 22 nanometre process in semiconductor manufacturing is not immediately visible for the computer user and from this perspective it may seem like yet another meaningless number in the computer specifications orat best a small incremental change. From the manufacturing perspective, on the other hand, shedding those extra nanometres has required enormous technological advances. One additional constraint onthe design is also the amount of heat being generated into a smaller and smaller space that still has to be dissipated through a cross-section of comparable size, leading to the invention of 'darksilicon' (powering down parts of a chip that are unused at any given moment). And on the other hand, the cumulative effect of such changes in terms of processing power, storage capacity and so on (Moore's lawand all that) enable new approaches to all kinds of problems thatwould have been quite impossible a decade or two ago. After all, eventoday's smartphones are more powerful than supercomputers in most ofthe 1980s.

The ability to deal with immense amounts of data in real time isdefinitely one of the two biggest driving forces for artificialintelligence in the foreseeable future. Recently I looked into the history of machine translation, and one of the earliest systems actually took twice as long just to do what amounts to looking up each individual word in a dictionary and stringing them together into an approximation of a translation as what it would take for a human translator to produce a correct translation. A system like Google Translate, on the other hand, has ginormous collections of multilingual documents with aligned language elements and uses them together with some heavy statistical processing to do the same job and produce at least something understandable if not correct in justfractions of a second. Oh, and the other driving force? Bio-inspired AI, or seeing how Nature has solved a given problem and trying to reproduce that in an artificial design.

As for my own work, I try to keep these different perspectives in mind, and while I occasionally like to throw words like 'robot judge' around, it is more as an abstract target (and of course also asprovocation) rather than as something I am actually concretely interested in implementing. But it is certainly helpful in trying to keep in mind the whole range of issues potentially involved in working with legal AI, and not just the issues du jour the research community finds interesting at the moment. In my opinion, one part of the problem is also that mainstream legal theory does not study law and legal reasoning in particular as a form of cognitive activity and has managed to all but ignore all the scientific progress made in both linguistics and psychology over the past fifty years, and in much of my work I take theories from those disciplines and try to apply them to questions of legal theory in a very general sense (mostly because almost nobody else seems to be doing it). Still, the best way forward for me seems to be trying to model some very small corner of the legal system using some particular technique (and I think I'm stuck with fuzzy logic at least until I've finished my dissertation) to see whether it works and then trying to see whether there are some broader conclusions to be drawn based on it. And in the best case it might even do something useful (read: marketable) at the same time. In this respect, AI & law seems to be about twenty years behind language technology.

We don't particularly need robot judges, but for example judicial decision-support systems could help actual judges in making correct and consistent decisions more efficiently and reliably, thus perhaps enabling them to spend more time on cases where uniquely human capabilities are really required. And if at the same time technology also revolutionizes the way legal services are provided (as predicted by Richard Susskind in particular), maybe the parties do not even have to go to court in the first place.

Augmenting Man

"The combination of machines and ICT has brought exponential development into the engineering world. As a result we see the emergence of autonomous machines. The complexity of tasks as well as the complexity of environments where these machines can work is steadily increasing. We can build cars that can navigate autonomously through city traffic finding destinations without any human intervention. Therefore, it is fair to say that machines have, on a functional level, already reached cognitive abilities comparable to horses or dogs. But this is not the end of development. Soon there will be no type of manual labor in which machines will not outperform humans. This is technically already true today. Currently, machines are merely held back by economic and societal constraints. The weakest of these constraints is the cost of hardware. Moore’s Law guarantees that computing power that today can only be found in supercomputers will be available in pocket sized devices in little a more than a decade. Some other constraints are more difficult to overcome. The more powerful and more complex a machine is, the more damage it can potentially create. This is the reason there are no self-steering cars on the roads yet. We have suggested a path of best practices and ethics to improve machines and reduce intentionally malicious behavior. Nevertheless, even those best practices leave us with a residual risk. This residual risk is not necessarily small. It may indeed be so large that certain types of machines will not be able to enter the market because of liability concerns. This limitation will only be overcome by the creation of an ultimate machine. For human parents responsibility and liability for a child ends with it becoming an adult. Similarly a machine can become an ultimate machine by emancipating itself from its manufacturer/owner and indeed becoming a distinct legal or even social entity. Interestingly, this can be done by creating a legal construct around this ultimate machine that in itself has economical value.

Nevertheless, the big question remains: how will our societies hold up to this rapid change? For example, currently our entire tax and social system, indeed most of our culture, is centered on the concept of work as the means of creating one’s livelihood. For example, the European Union has set a goal of increasing the part of the population (between 15 and 64 years of age) in gainful employment to 70 per cent up to 2010. Yet when machines are able to perform manual labor cheaper and more efficiently than humans, what jobs will remain? Former US Secretary of Labor, Robert Reich, assumes that manual labor will eventually be replaced completely by machines. Nevertheless he argues that there will still be a high demand for a human work force. These new workers will have to be highly educated and trained “symbolic analysts” – lawyers, doctors, journalists, consultants, and the like – which create value beyond mere manufacturing. However currently only a fraction of the labor force is capable of performing these jobs. Even though goverments have stated their intention to increase investment in education it is questionable whether this goal can be achieved for everyone. And even if it were possible, the advancement in information technology is not restricted to manual labor. Machines have augmented the physical performance of man to the point were he becomes superfluos. The same augmentation is also taking place with our cognitive abilities. The famous quote of the computer being a “bicycle for the mind” becomes evident when we consider the vast amount of data a single person can analyze with the help of a personal computer. Therefore, the observation that machines in the long run are not destroying jobs but creating new ones is merely that; an observation and not a law. There might well be a threshold of automation that changes the rules of the game entirely. If that should happen this would be one aspect in which we have to change our culture radically. In any case, how well we are prepared for these new machines will determin the social acceptance and ultimately the cost of the transition. Since development is still gradual, there will be several years left to create new practises. There is likely not a simple nor a single answer. The convergence of disciplines and the accelerating speed of technological progress will require a holistic approach and result in ad-hoc solutions. Fortunately, we can start learning about the problem and its solutions already today. After all, the future is already here, just not equally distributed."

William Brace, Anniina Huttunen, Vesa Kantola, Jakke Kulovesi, Lorenz Lechner, Kari Silvennoinen and Jukka Manner,
"Augmenting Man,"
in Bit Bang, Rays to the Future, Yrjö Neuvo and Sami Ylönen, Helsinki: Helsinki University Print, 2009, p. 236-263

File sharing + Robots = "Low Orbit Server Drones"

Have you heard about the Pirate Bay's latest idea? They are planning Low Orbit Server Stations (LOSS). In other words, they would like to host "parts of their site in GPS-controlled drones, instead of old-fashioned data centers." Due to a court decision my operator Elisa has blocked its customers access to the Pirate Bay. Consequently, I am not able to check what they say about the subject, but below you can find some information:

The Pirate Bay Attacks Censorship With Low Orbit Server Drone

The Pirate Bay Planning "Low Orbit Server Drones"

And I thought that there is no way to combine file sharing and robot studies...