4 ms·

What a silly article. PageRank is based on eigenvector centrality[1], which is one of a number of well-understood centrality measures that have been around for

by frisco 13y ago

What a silly article. PageRank is based on eigenvector centrality[1], which is one of a number of well-understood centrality measures that have been around for a long time. Google's innovation was noticing the importance of the link structure of the internet and knowing to apply EC to it - not inventing the graph theory underlying EC. Edit: and knowing to apply EC to that graph structure. [1] http://en.wikipedia.org/wiki/Centrality http://en.wikipedia.org/wiki/Centrality

nullc 13y ago
What they patented, however, is more broad than that still: 1. A computer implemented method of scoring a plurality of linked documents, comprising: obtaining a plurality of documents, at least some of the documents being linked documents, at least some of the documents being linking documents, and at least some of the documents being both linked documents and linking documents, each of the linked documents being pointed to by a link in one or more of the linking documents; assigning a score to each of the linked documents based on scores of the one or more linking documents and processing the linked documents according to their scores.
- pudquick 13y ago
  .. And yet still not invalidated by this prior art (if you were to stretch the concept) due to key phrases like 'computer implemented method'. Computers in the 1940s were things like ENIAC and more likely to be involved in computing firing tables or cryptanalysis. Now - if you were to look at Kleinberg's HITS instead you may possibly have a better argument, but I'm betting Kleinberg's prioritization method is still technically different enough than PageRank for that patent to stand. Given that Brin and Page actually reference the paper, I'm sure they've had patent lawyers make sure it wasn't infringing.
- 6ren 13y ago
  Patent claims usually begin with the broadest claim possible, then narrow it in following claims. If the first is found invalid, the other claims may still succeed. Thus, it makes no sense for the first claim to be certain to stand on its own; strategically, it should be over-broad. From a computer science/mathematics point of view, this is also a convenient way to factor out commonalities of several claims, instead of duplicating them ad nauseam. For example, in the PageRank patent, claim 1 is immediately followed by six dependent claims giving further detail on it (as do ten other claims scattered through the rest): https://www.google.com/patents/US6285999 https://www.google.com/patents/US6285999 Thus, it is inaccurate to treat the first claim as "what they patented".
  - throwawaykf02 13y ago
    While it's true that the independent claims try to be as broad as possible and dependent claims narrow them down, each claim is still evaluated -- and must be valid -- on an individual basis. And if infringement of a patent is to be determined, you'd start with the independent claims and work your way down, as those have the broadest scope. So, yes, it does make sense to say that the patent claims what nullc quoted. It's only slightly inaccurate in that the patent also claims a lot more, but the independent claims define the broadest scope.
    - 6ren 13y ago
      Oh dear, you and nullc are right, and I am wrong. My apologies. This patent was granted, and so they do have a patent for claim 1 (and also for the other claims). I was confusing it with a patent application. It's surprising to me that even the cursory prior art search of examination didn't turn up this idea, since it's so simple, doesn't use any of the meat of the invention described, and the idea of ranking by citation was well-known, e.g. for academic papers. I think applying that to hypertext links is clever and insightful... but not patent-worthy (like many other patents). Upon litigation, I think claim 1 might not stand. But regardless, the patent was granted, so they do have a patent on claim 1. Thanks for taking the time to correct me.
mayank 13y ago
> Google's innovation was noticing the importance of the link structure of the internet - not inventing the graph theory underlying EC. Actually, I'd say that Google's innovation was applying the eigenvector centrality computation to the Web graph, not discovering any particularly meaningful property (the "random surfer") of the Web graph. I'd call it an extremely skillful application of an old algorithm to a new problem domain.
spullara 13y ago
We shouldn't allow patents on old X applied to new Y domain.
- gfodor 13y ago
  You just gave a fairly abstract definition of "innovation."
  - walid 13y ago
    Nope, @spullara just gave a definition of prior art :|
    - gfodor 13y ago
      The point being the difference between "innovation" and "prior art" is a tricky one, since all innovation builds upon previous knowledge of some kind.
floobynewb 13y ago
Damn you beat me to it.
wicknicks 13y ago
Exactly. Page and Brin cite some papers (or atleast Kleinberg's paper does -- which inspired Pagerank) which deal with ranking authors based on their citation graph. This article makes no sense whatsoever.

by frisco 13y ago

nullc 13y ago

What they patented, however, is more broad than that still: 1. A computer implemented method of scoring a plurality of linked documents, comprising: obtaining a plurality of documents, at least some of the documents being linked documents, at least some of the documents being linking documents, and at least some of the documents being both linked documents and linking documents, each of the linked documents being pointed to by a link in one or more of the linking documents; assigning a score to each of the linked documents based on scores of the one or more linking documents and processing the linked documents according to their scores.

pudquick 13y ago
.. And yet still not invalidated by this prior art (if you were to stretch the concept) due to key phrases like 'computer implemented method'. Computers in the 1940s were things like ENIAC and more likely to be involved in computing firing tables or cryptanalysis. Now - if you were to look at Kleinberg's HITS instead you may possibly have a better argument, but I'm betting Kleinberg's prioritization method is still technically different enough than PageRank for that patent to stand. Given that Brin and Page actually reference the paper, I'm sure they've had patent lawyers make sure it wasn't infringing.
6ren 13y ago
Patent claims usually begin with the broadest claim possible, then narrow it in following claims. If the first is found invalid, the other claims may still succeed. Thus, it makes no sense for the first claim to be certain to stand on its own; strategically, it should be over-broad. From a computer science/mathematics point of view, this is also a convenient way to factor out commonalities of several claims, instead of duplicating them ad nauseam. For example, in the PageRank patent, claim 1 is immediately followed by six dependent claims giving further detail on it (as do ten other claims scattered through the rest): https://www.google.com/patents/US6285999 https://www.google.com/patents/US6285999 Thus, it is inaccurate to treat the first claim as "what they patented".
- throwawaykf02 13y ago
  While it's true that the independent claims try to be as broad as possible and dependent claims narrow them down, each claim is still evaluated -- and must be valid -- on an individual basis. And if infringement of a patent is to be determined, you'd start with the independent claims and work your way down, as those have the broadest scope. So, yes, it does make sense to say that the patent claims what nullc quoted. It's only slightly inaccurate in that the patent also claims a lot more, but the independent claims define the broadest scope.
  - 6ren 13y ago
    Oh dear, you and nullc are right, and I am wrong. My apologies. This patent was granted, and so they do have a patent for claim 1 (and also for the other claims). I was confusing it with a patent application. It's surprising to me that even the cursory prior art search of examination didn't turn up this idea, since it's so simple, doesn't use any of the meat of the invention described, and the idea of ranking by citation was well-known, e.g. for academic papers. I think applying that to hypertext links is clever and insightful... but not patent-worthy (like many other patents). Upon litigation, I think claim 1 might not stand. But regardless, the patent was granted, so they do have a patent on claim 1. Thanks for taking the time to correct me.

mayank 13y ago

> Google's innovation was noticing the importance of the link structure of the internet - not inventing the graph theory underlying EC. Actually, I'd say that Google's innovation was applying the eigenvector centrality computation to the Web graph, not discovering any particularly meaningful property (the "random surfer") of the Web graph. I'd call it an extremely skillful application of an old algorithm to a new problem domain.

spullara 13y ago

We shouldn't allow patents on old X applied to new Y domain.

gfodor 13y ago
You just gave a fairly abstract definition of "innovation."
- walid 13y ago
  Nope, @spullara just gave a definition of prior art :|
  - gfodor 13y ago
    The point being the difference between "innovation" and "prior art" is a tricky one, since all innovation builds upon previous knowledge of some kind.

floobynewb 13y ago

Damn you beat me to it.

wicknicks 13y ago

Exactly. Page and Brin cite some papers (or atleast Kleinberg's paper does -- which inspired Pagerank) which deal with ranking authors based on their citation graph. This article makes no sense whatsoever.