News for hackers. Stuff that matters.
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nkurz on Tuesday September 07, 2010 @02:16AM (#1668242)
This is a good layman's introduction to modern search techniques, but to someone not in the SEO field it feels like a very strange inversion of priorities. To me, like most people, the surprise is how effective techniques like LDA[1] can be in characterizing a document, but the 'surprise' in the article is that LDA correlates to Google search order better than a more simplistic model.

To a technologically savvy but naive outsider, this might seem obvious: shouldn't pages that rank highly in Google have strong topic-based correlation to pages that the user wants to see? But from the SEO perspective, I guess the conclusion would be that your page is more likely to be ranked highly if it includes all the trappings of other high ranked pages, with, you know, like synonyms and stuff. At a certain point, one has to start thinking, wouldn't it be simpler to make a page that people actually want to find?

Are there good examples of actually useful pages that Google doesn't do a good job of ranking? I occasionally find myself lately getting frustrated with Google about ignoring my rarer search terms, but generally I find the good pages are at the top if they exist at all.

[1] LDA is Latent Dirichlet Allocation, which is very similar to Latent Semantic Analysis, which in turn is very similar to Principle Component Analysis and Singular Value Decomposition. So it's possible you've already heard of the concept, but coming from another angle in another field.

Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nodata on Tuesday September 07, 2010 @08:44AM (#1668722)
> Are there good examples of actually useful pages that Google doesn't do a good job of ranking?

Yes. Reviews by actual people.

A long, long time ago (last year I think) I used to use google blogsearch to get at the "other side" of hardware reviews: what people who are not being paid to review hardware think of a specific piece of hardware.

Doesn't work any more. There's almost nothing but spam. It's hopeless.

Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nkurz on Tuesday September 07, 2010 @02:48PM (#1669712)
I agree. There are definitely cases where spam has displaced good results. Although poorly phrased, I guess I was asking a different question: apart from intentional attempts to game the system, are there cases where pages one wants to see are consistently ranked after pages one does not? Which is to say, are there improvements that could be made to Google's ranking algorithms beyond making it more spam resistant?
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nl on Tuesday September 07, 2010 @02:21AM (#1668251)
I though it was kind of odd they did LDA rather than something more broadly used (eg LSA).

But I've never really looked at LDA, and Wikipedia says: Compared to standard latent semantic analysis which stems from linear algebra and downsizes the occurrence tables (usually via a singular value decomposition), probabilistic latent semantic analysis is based on a mixture decomposition derived from a latent class model. This results in a more principled approach which has a solid foundation in statistics. so maybe they made the right choice. (Not that I see what "solid foundation in statistics" really means in this context)

Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by cdavid on Tuesday September 07, 2010 @03:58AM (#1668364)
LSA is relatively similar in some abstract sense to the published pagerank algorithm. LDA is more powerful, in the sense that it can account for more complex relationships (but may be less accurate with large number of data - I have really no idea how those would scale and compare at google-like size).
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nl on Tuesday September 07, 2010 @07:08AM (#1668568)
Can you explain this some more?

My understanding of LDA is that it gives you document scores against queries based on the topics extracted using the LDA algorithm on the text in the page.

Pagerank, on the other hand scores based on external pointers (ie, references) but doesn't have anything to do with the text on the page.

Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by noelwelsh on Tuesday September 07, 2010 @09:22AM (#1668794)
It's the abstract sense that is important. Pagerank is a dimensionality reduction technique. It finds the first eigenvector of the transition matrix. Eigenvectors = PCA. LSI is basically PCA, but applied to the document-term matrix. LDA is a dimensionality reduction technique that makes use of more information.
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by nl on Wednesday September 08, 2010 @07:00PM (#1673817)
Oh, I see.

I thought you were talking about some functional similarities, not the mathematical similarities.

Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by yread on Tuesday September 07, 2010 @04:25AM (#1668397)
For some applications Dirichlet mixtures are even better. I used this tool, which I can recommend, for deriving the parameters of the mixtures http://chasen.org/~daiti-m/dist/dm/
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by jkan on Tuesday September 07, 2010 @05:05PM (#1670091)
I think "solid foundation in statistics" refers to the fact that LDA has a proper generative model of documents and topics, while LSA doesn't -- so that it's easier to reason about LDA and to build more complex models on top of it.
Re:Latent Dirichlet Allocation Surprisingly Well Correlated ... by equark on Tuesday September 07, 2010 @03:18AM (#1668327)
It means there is a likelihood function.