Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google's PageRank patent has expired (2019)

patents.google.com

Post reply on HN
Print view View original

81–90 of 173 posts

Re: Google's PageRank patent has expired (2019)

#81

Post by srean »

samhw wrote: post #66
srean wrote: post #56

Earlier quoted context omitted.

The HITS algorithm has one drawback that it is super easy to game it. PageRank is not entirely resistant, but its a little more robust. For an algorithm that is used in the real world where commercial interests rules, sensitivity to adversarial attacks is very important -- Game Theory inspired algorithms will not be out of place. Note, one can definitely adopt 'hardening' strategies for both HITS and PageRank. Even a…

> The HITS algorithm has one drawback that it is super easy to game it. PageRank is not entirely resistant, but its a little more robust. > Create a harvester page that points to lots and lots of popular, high traffic pages on the internet. By virtue of doing this it can accumulate a lot of Hubs score which it can redirect as an Authority score to an intended page. I'm not sure how HITS is any more "easy to game" tha…

One could pose this as an adversarial game. For the simplistic case consider two participants -- (i) the ranker that chooses a ranking scheme (we need to constrain the space of ranking schemes somehow for this to lead to any useful formulation), (ii) web page who tries to outrank other pages by strategically linking to other pages, and possibly buying links to itself from other pages. One can give (ii) a budget to add and delete links and pages that it can control. In this framework one then try to compute what's an equilibrium strategy. The multiplayer version is a lot more complicated.

If you check my original comment, I gave a simple scheme to attack HITS rank. The main drawback is that one can 'harvest' Authority score using 'out-links'. Outlinks are cheap and easy, compared to 'inlinks'. Sybil attack is a little harder for Pagerank.

Re: Google's PageRank patent has expired (2019)

#82

Post by thethimble »

s17n wrote: post #60
creamynebula wrote: post #32

Earlier quoted context omitted.

My search results were a lot better in 2006 when, I assume, they didn't have all these ML pipelines...

That's because the web was better in 2006. I think Google has made one change for the worse, though, which is strongly favoring more recent content. Increasingly, I think that change has been a big contributor to the decay of the web since.

I think you’ve forgot the websites with walls of white-on-white keywords/links at the bottom, trying to game pagerank. Many of these models were marked improvements compared to the generations before.

Re: Google's PageRank patent has expired (2019)

#83

Post by photochemsyn »

losvedir wrote: post #51
creamynebula wrote: post #32

Earlier quoted context omitted.

My search results were a lot better in 2006 when, I assume, they didn't have all these ML pipelines...

> My search results were a lot better in 2006 when, I assume, they didn't have all these ML pipelines... That's like an old person complaining that their body felt a lot better back in 2006 when, I assume, they didn't have to use their walker and glasses all the time...

I find that Google search results are still the best of any search engine for specific computer science and other specific tech-related topics, as long as you construct a fairly complex search string. However, for general information on things like world events, local news, national politics, etc. it's become little more than a mirror for corporate and state propaganda outlets. This is likely due to those very ML pipelines mentioned above:

Vince (authority/brand power), Panda (inbound link quality), Penguin (content quality)...

This represents a pretty severe narrowing of results on information and opinions - possibly the worst results are on Youtube searches for newsworthy events. It'd be very interesting to see what kind of content a pure PageRank algorithm-based search engine would generate today, and I'd be very interested in using such a search engine. Now, would it be overrun by SEO? I don't know, but it'd be worth finding out.

I kind of wonder if Google Scholar is purely PageRank or citation-count based, it still gives very useful results with relatively simple query strings.

Re: Google's PageRank patent has expired (2019)

#84

Post by not2b »

charcircuit wrote: post #62
heretogetout wrote: post #20

Earlier quoted context omitted.

Quotes aren't what they once were. They don't guarantee that the term you quoted will be in the results.

>They don't guarantee that the term you quoted will be in the results. They do guarantee it. Sometimes Google is buggy and messes up in determining what's page is actually visible on the page compared to just being somewhere in the HTML.

The page Google saw might not be the one you get if you click the link they give you. This kind of thing is in the SEO bag of tricks.

Re: Google's PageRank patent has expired (2019)

#85

Post by NateEag »

bogwog wrote: post #73
scroot wrote: post #17

Cool. Now let's get the Library of Congress or NARA -- which are natural homes for this kind of thing -- to implement the public option free of profit-driven cruft.

That would actually be pretty cool. Too bad government software has a tendency to come out shitty and expensive.

If the government committed to the project being OSS, they might be able to recruit some reasonably skilled programmers.

Re: Google's PageRank patent has expired (2019)

#86

Post by samhw »

srean wrote: post #81
samhw wrote: post #66

Earlier quoted context omitted.

> The HITS algorithm has one drawback that it is super easy to game it. PageRank is not entirely resistant, but its a little more robust. > Create a harvester page that points to lots and lots of popular, high traffic pages on the internet. By virtue of doing this it can accumulate a lot of Hubs score which it can redirect as an Authority score to an intended page. I'm not sure how HITS is any more "easy to game" tha…

One could pose this as an adversarial game. For the simplistic case consider two participants -- (i) the ranker that chooses a ranking scheme (we need to constrain the space of ranking schemes somehow for this to lead to any useful formulation), (ii) web page who tries to outrank other pages by strategically linking to other pages, and possibly buying links to itself from other pages. One can give (ii) a budget to ad…

> If you check my original comment I gave a simple scheme to attack HITS rank. Sybil attack is a little harder for Pagerank.

OK, but how is it harder for PageRank? I can't really see any differences in the semantics of the two algorithms, so I'm not sure what kind of added vulnerability one or the other could have.

> One could pose this as an adversarial game.

Yeah, I appreciate that, that's what I was referring to as "the idea that people will try to game it". It's not really the kind of 'game' that would be considered in game theory, though, because it doesn't have any interesting or emergent properties - the designer's response will just be "oh yeah we should stop people gaming our algorithm".

Re: Google's PageRank patent has expired (2019)

#87

Post by anotherman554 »

s1k3s wrote: post #13
tyingq wrote: post #5

I'm not sure how useful this might be to anyone anymore. Google still uses it for the base ranking. But, the results are then run through a variety of add-on ML pipelines, like Vince (authority/brand power), Panda (inbound link quality), Penguin (content quality), and many others that target other attributes (page layout, ad placement, etc). Then there's also more granular weightings for things like "power within a n…

> "power within a niche" In my country we have a website that has complete monopoly over sales of used items, services etc. I'm always amazed that Google is able to determine and put it in top 3 results for a wide majority of searches. "laptop mouse" brings up that site, but "kitchen sink" doesn’t, presumably because people don’t buy used sinks.

If google returned the used item site for "kitchen sink" and nobody clicked on that result, it would be easy for an algorithm to make the connection that this is a bad result that shouldn't be ranked very high.

Re: Google's PageRank patent has expired (2019)

#88

Post by heretogetout »

charcircuit wrote: post #62
heretogetout wrote: post #20

Earlier quoted context omitted.

Quotes aren't what they once were. They don't guarantee that the term you quoted will be in the results.

>They don't guarantee that the term you quoted will be in the results. They do guarantee it. Sometimes Google is buggy and messes up in determining what's page is actually visible on the page compared to just being somewhere in the HTML.

That's a loose definition of guarantee, I think. They know what's on the page because they include a brief summary in the results, but even there quoted terms don't always appear.

Here's an example someone posted a few months back where google decided the user didn't really want what they said they wanted:

https://news.ycombinator.com/item?id=30132344

Re: Google's PageRank patent has expired (2019)

#89

Post by ivanbakel »

raunak wrote: post #75
losvedir wrote: post #51

Earlier quoted context omitted.

> My search results were a lot better in 2006 when, I assume, they didn't have all these ML pipelines... That's like an old person complaining that their body felt a lot better back in 2006 when, I assume, they didn't have to use their walker and glasses all the time...

Search is objectively worse today than it used to be, it's not just that "it's harder to use for old people", it's just worse.

>Search is objectively worse today than it used to be

How would you demonstrate that search is objectively worse? And how would you then show that it's a result of Google's algorithms, and not a consequence of the content of the Internet changing significantly?

Re: Google's PageRank patent has expired (2019)

#90

Post by tyingq »

ricardo81 wrote: post #52
tyingq wrote: post #36

Earlier quoted context omitted.

My view of it is that they basically lost the SEO spam wars. Without the ML pipelines, the top results would all be dominated by the smaller, highly skilled, SEO manipulators. They didn't find a way to cleanly excise that spam, so they resorted to a very imperfect hammer...giving a lot of SEO weight to large corporate entities. So basically, a different kind of spam dominates now.

>smaller Perhaps in other words, market forces (I'm thinking ROI) and economies of scale will win.

What I was talking about is that there was once a pretty large community of smaller SEO manipulators that did a lot of experimentation on what really worked. Especially in the "black hat" realm. Finding things like areas on big-brand sites that would accept user-generated content, comments, etc, where you could bypass some checks and insert links to sites.

They would experiment in a pretty deep way, varying things like the rate of new links, type of new links, variations of anchor text / bare links, and so on.

It USED to be very effective.

The larger entities don't really have to be that detailed. If you have that brand power, you can just cut partnership/cross-link deals and pay a little attention to things like anchor text in links, contextual text around the link, etc.

Edit: Ah, yeah, agreed. What got lost in all this was good content that had no big brand behind it. The indiscriminate hammer Google used to kill off small-guy SEO spam also pushed a lot of actually good stuff, stuff that never did any SEO at all, off the first page.

Post reply on HN