23 October 2016

If the Patent Office is Correct, Google’s Seminal ‘PageRank’ Invention Would Be Unpatentable in Australia

SearchOn 4 September 1998 a company was founded in order to commercialise some ideas that were being developed by two PhD students at Stanford University.  The most important of those original ideas had resulted in the filing of a US provisional patent application on 10 January 1997, in the name of one of those two PhD students.  This was followed by a non-provisional application, filed on 9 January 1998, which went on to become US patent no. 6,285,999.  Ownership of the patent remained with Stanford, which licensed it exclusively to the students’ company in exchange for 1.8 million shares. 

In 2005, those shares were sold for US$336 million.

Today, a few hundred million dollars is small change to those former PhD students, and the company that they founded.  The inventor on the original patent application was Lawrence (‘Larry’) Page, and his fellow student/founder was Sergei Brin.  The invention was a computer-implemented method for determining the ‘importance’ of web pages, for ranking purposes, which became known as PageRank (a play on the inventor’s name, and the purpose of the invention).  And the little company they started was, of course, Google, Inc.

While the original PageRank application has spawned a number of further US patents, it was never filed anywhere else in the world.  There was therefore never any possibility of an Australian patent being granted.  It is nonetheless concerning that, if a recent decision by a Delegate of the Commissioner of Patents, Amadeus S.A.S [2016] APO 71, is correct, the invention that helped launch one of the biggest companies the world has ever seen would be ineligible for patenting in Australia.

PageRank – an Idea that Launched a Giant

Prior to Google, finding information on the Web was very much a hit-and-miss affair.  Unlike conventional databases (e.g. library catalogues, indexes of scholarly articles, or even the collected contents of published patent applications) the Web was (and still is) a completely disorganised collection of random and diverse information.  Anybody can add to it simply by starting their own web site.  The quality of content varies enormously, from the spectacularly informative to the dangerously inaccurate.  Most web sites – let alone individual web pages – are not subject to any form of editorial review, categorisation or classification.  The Web may be the greatest source of information ever created by humanity, but it is also the greatest source of drivel, lies and lunacy.  (I mean, take this blog... please!)

Some of us remember web search engines before Google.  Most of them, and the companies that created them, are dead, dying, or have pivoted to some other application of the technology and/or brand recognition they created.  These include WebCrawler, Lycos, HotBot, Northern Light and Ask Jeeves, along with AltaVista, which was taken over by Yahoo! in 2003 (and it should be noted that, pre-Google, Yahoo! provided a web directory service, but was not a true search engine).

I used all of these search engines, and I still clearly remember the ‘Google revolution’.  I was working in academia at the time and, along with many of my colleagues, was on to it relatively early, in around 1998-9.  By 2000, however, Google was becoming widely-recognised by a broad cross-section of web users as a clearly superior search engine.  It was simple to use, having only a single search box and no need to learn ‘advanced search’ features to obtain useful results.  Almost any remotely sensible search term would produce something of relevance.  Most importantly, for the first time the most useful, reliable and relevant results were actually very likely to appear on the first page. 

It was Google’s superior ranking ability that really got it off the ground and built the initial traction that ultimately enabled it to gain, and maintain, clear market leadership in web search.  Of course, Google had to do very much more than this to become a successful business, and had to continue to innovate in order to build on its initial competitive edge.  But, nonetheless, Google is a textbook example of a company that was able to bootstrap off a key piece of unique intellectual property, namely the patented PageRank algorithm.

The PageRank Concept

The idea behind PageRank is relatively simple.  It had already been recognised that if a web site, or individual page, was important (in the sense of being reliable, useful, or otherwise worthy of reference), it is likely that a number of other sites and pages will link to it.  However, the converse is not true.  Just because a site has a large number of incoming links does not necessarily mean that it is important.  It is, for example, easy enough to obtain incoming links by setting up a number of separate sites specifically for this purpose, or by buying them from the owners of other sites who are willing to add links in exchange for money or other valuable consideration.

Page’s insight was that links from important sites should be more indicative of the importance of the target site.  Thus PageRank assigns weightings to incoming links based upon the importance (i.e. the PageRank score) of the site/page providing each link.  A link from a page that itself has a high PageRank contributes more than a link from a page with a lower PageRank score.  Contrived, or otherwise low-value, pages therefore contribute very little to the PageRank score of the target page.

Implementing PageRank in practice is not as simple as the above description of the general concept behind it.  The Web has no fixed beginning or end.  It contains a tangle of links, back-links and link cycles (i.e. sequences of links that ultimately lead back to the original page).  Determining the PageRank score of a page requires knowledge of the PageRank scores of all the pages that link to it, which in turn require knowledge of the PageRank scores of all the pages that link to those pages... and so on, ad infinitum (or nearly so, given the ever-expanding size of the Web).  And the Web is dynamic – pages are edited, added and deleted constantly.  All of this complicates the calculation process.

The PageRank Patent Specification

As patent specifications go, that of US patent no. 6,285,999 is relatively concise.  It includes just three simple drawings, two of which illustrate links between pages on the Web, with the third being a very simple flowchart (three linearly-connected boxes) illustrating the PageRank algorithm.  The description is just under four pages long, with a further one-and-a-half pages being occupied by the patent’s 29 claims.

The majority of the description in the patent specification the specification is directed to the mathematical equations and general numerical methods (e.g. iterative recalculation of ranks to converge upon ever more accurate approximations of actual rank value) that may be implemented in order to compute PageRank scores for every page on the Web.  Thus, a mathematical solution to the problems of the prior art – namely, the unreliability of simply counting incoming links regardless of the significance of the linking page – is presented in the specification.

The invention disclosed in the patent specification thus enables an improved ranking of search results to be presented to the user.

However, the specification does not identify any particular further ingenuity or technological advance that must be made in order to put the invention into effect using a computer.  There is, indeed, no description of any code or specific programming required to implement the PageRank method within a search engine.  The computer implementation of the method is left to the reader, who (as a competent computer programmer) is presumably expected to have no particular difficulty in translating the mathematical equations and numerical computation techniques outlined in the specification into working computer code.

The invention resides in the PageRank algorithm itself which, being implemented on a computer, results in the provision of an improved search engine.  This is a technical solution to a problem that arises in the field of computer and internet technology itself.  It is not a ‘business method’, nor merely an abstract idea or something that consists only of mental steps.  In every respect, it is the kind of thing that has long been considered patent-eligible in Australia – at least since the decision of the Federal Court in IBM v Commissioner of Patents [1991] FCA 625.

The Invention and Outcome in Amadeus S.A.S.

I do not intend to discuss the Amadeus patent specification and claims in detail, in part because, although I am no longer employed at Watermark, I was the attorney who prepared the submissions on behalf of Amadeus S.A.S., and it would therefore be inappropriate for me to go beyond the information that is available in the public record.  For present purposes, it largely suffices for me to quote from the Delegate’s decision.

Briefly, the invention in the Amadeus case relates to ranking the results of searches conducted in travel databases.  In particular, in the case of retrieving possible travel itineraries (e.g. flights) for corporate travellers, paid for by their employers, the inventive method takes into account preferences of both the traveller (e.g. for particular seating, airlines, travel times etc) and of the employer (e.g. cost constraints).  There is, however, much more to it than this simple characterisation suggests, as the decision explains (at [23]):

...the invention takes into account a combination of a value utility function and a price utility function by way of a weighted sum to define a composite utility function. It analyses actual selections of travel proposals made by a group of travellers, such as a group of employees of a particular corporate entity, in order to infer traveller preferences and thus define the value utility function. The price utility function is determined by presenting the payer with a number of different scenarios such that payer sensitivity data is obtained which is analysed to determine relevant parameters of the price utility function. Once a suitable composite utility function has been defined for a particular group of travellers and the corresponding payer, it is then used to rank the results returned to a user in response to a search query thereby simplifying the task of the user in identifying a preferred travel proposal that best meets the user's preferences.

The Delegate accepted (at [27]) that ‘[i]t is apparent that a mathematical solution to the problems of the prior art has been presented in the specification.’  Furthermore, he did not dispute the nature of the contribution made by the invention disclosed in the specification (at [28]):

Considering the claimed invention in the context of the specification as a whole and in light of the common general knowledge in the art, it is clear that claim 1 is directed to a computer implementation of a method for providing ranked travel proposals to a user.

Nor did the Delegate dispute that this was a new and useful contribution to the field (at [30]):

To the extent that newness and usefulness of the method within a domain specific search engine are concerned, I note that the lack of objections relating to novelty and usefulness of the invention in the second examination report confirm that this is also the case.

Nonetheless, the Delegate determined that the claimed invention was not patent-eligible under the Australian ‘manner of manufacture’ requirement (at [31-2]):

While asserting that the implementation of such as a method for a search engine self-evidently solves a technical problem within the computer system, the applicant has failed to identify any ingenuity or advances in technology that arise in the use of a computer.  It is noteworthy that technical details representing some ingenuity or advance in computer technology as such, or steps foreign to the normal use of a computer, are also completely absent from the specification. One therefore concludes that the computer implementation of the method has been left to the skilled addressee.

It is thus apparent that the substantive effect of the claimed invention lies in a mere scheme for providing ranked travel proposals to a user employing the computer for its well-known and understood functions wherein the computer is simply being used as an intermediary or a tool for putting the scheme into effect while adding nothing to the idea of providing ranked travel proposals to a user.  In my opinion, the substance of the claimed invention is not technical and does not involve the solution of a technical problem.  Therefore claim 1 is not for a manner of manufacture.

The Delegate went on to conclude not only that all of the other claims in the application also fail to define patent-eligible subject matter, but that there is nothing whatsoever in the entirety of the specification that could form the basis of a patentable claim (at [35]).  The application was therefore refused.

Where the Decision Went Awry

Notably, up until the point at which the Delegate raised the issue of a lack of any description of ingenuity or advance in computer technology, all signs would have pointed to the invention being patent-eligible.  However the logic here, i.e. that it is ‘apparent that the substantive effect of the claimed invention lies in a mere scheme’ because ‘the computer implementation of the method has been left to the skilled addressee’, puts the cart before the horse. 

It is clear from the relevant authorities that it is necessary first to examine a claimed invention ‘to ascertain whether it is in substance a scheme or plan or whether it can broadly be described as an improvement in computer technology’: Commissioner of Patents v RPL Central Pty Ltd [2015] FCAFC 177 at [96].  Where the invention is in substance a mere scheme, which ‘is not, per se, patentable’, only then does it become important to look further in order to determine whether there is some patentable ingenuity present in the computer implementation itself.

Accordingly, once the Delegate had determined that ‘it is clear that claim 1 is directed to a computer implementation of a method for providing ranked travel proposals to a user’, that this method was new and useful, and that it provided a ‘solution to the problems of the prior art’, this should have been the end of the enquiry.  An improved method for ranking computer search results, whether from a database of travel schedules or from any other database (e.g. a corpus of web documents) is not a mere scheme.  It is, of its very nature, an improvement in computer technology.

Conclusion – Confusion and Inconsistency Reigns

I no longer have clients, or the reputation of an employer, to consider, so I am just going to come straight out and say it – this is a terrible and flawed decision.  It is evident that if the phrase ‘ranked travel proposals’ in the Delegate’s reasoning is replaced with ‘ranked web search results’ it maps completely and consistently onto the description and claims of the Google PageRank patent.

Accordingly, if this decision were correct, an invention that launched a multi-multi-billion-dollar company, and returned hundreds of millions of dollars to the academic institution that incubated that invention, is not patent-eligible in Australia.  I was under the impression, from before I entered the patent profession 14 or so years ago, that this was precisely the kind of innovation and commercialisation activity that the patent system was supposed to foster, in all fields of technology.

At the same time, recall that less than three months earlier a different Delegate of the Commissioner determined that the expedient of combining the two functions of game selection and bet denomination selection into a single touch-screen button on a multi-game casino gaming machine does constitute patent-eligible subject matter, notwithstanding that only ‘generic computer implementation’ was required.

So, let us just put this in perspective.  Based on the reasoning in these two decisions:
  1. quantum leap in web search technology – not eligible;
  2. enabling people to more efficiently gamble away their money by streamlining a user interface – eligible. 
Welcome to the wonderfully confusing and inconsistent world of the Australian Patent Office, folks.  And get used to it, because I do not see it changing any time soon without some concerted intervention.

Personally, I think the Commissioner of Patents needs to hold a crisis meeting.  She should get all of the hearing officers who deal with computer-implemented technology, and any other relevant senior Patent Office staff, into a room to work out exactly what the Office’s position is on this kind of subject matter.  Ideally, the Commissioner would publish a draft practice note and accept submissions from stakeholders before finalising a consistent and comprehensible approach to the assessment of claims directed to computer-implemented inventions, as she did last year for genetic technologies in the wake of the High Court’s decision in D’Arcy v Myriad Genetics Inc [2015] HCA 35.

Stakeholders – whether they be applicants for, or opponents of, patents on computer-implemented inventions – deserve no less.  The current uncertainty and inconsistency is the greater of all possible evils.


Post a Comment

Copyright © 2014
Creative Commons License
The Patentology Blog by Dr Mark A Summerfield is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Australia License.