Thursday, September 29, 2005

Intrinsic analysis vs social smarts

Last.fm and Pandora take two different approaches to the problem of helping people find new music that they are likely to enjoy.

Pandora asks you to provide the name of an artist or song as a starting point. Then, based on its prior analysis of the song or artist, it recommends 'similar' songs that meet the search criteria. Amazingly, the analysis of each song is perfromed by humans

Together our team of thirty musician-analysts have been listening to music, one song at a time, studying and collecting literally hundreds of musical details on every song. It takes 20-30 minutes per song to capture all of the little details that give each recording its magical sound - melody, harmony, instrumentation, rhythm, vocals, lyrics ... and more - close to 400 attributes!


The Pandora player streams these relevant songs and allows you to customize with your likes/dislikes. So, you can modify the default recommendation playlist but, out of the box, the default is not targetted at particular users.



Last.fm bases its recommendations as to what you might enjoy hearing based not on any intrinsic attributes of a song or artist, but rather on the proximity of songs to each other in a social space. For instance, if two users share a taste in a particular artist (these tastes tracked by a music plug-in for players that pushes played songs to the Last.fm database and presumably weighted by the number of plays) then the sytem assumes that one user would enjoy listening to other songs in the other user's playlist. What I hear is based partly on what those who are close to me in this "taste space" like to listen to. A refinement of the Last.fm model would allow me to assing greater weight to particular users (Last.fm allows me to identify 'friends' but it's not clear whether they are factored into the recommendation algorythm).

Early comparisons of the success rate, (e.g. delivering me a song that I do indeed enjoy) seems to favour Pandora, perhaps not surprising given the heavy-lifting that has gone into creation of the database. Last.fm's recommendations will likely improve as more and more users join-up so that the current occasional perversions (see below) would be swamped out by numbers.

Saturday, September 24, 2005

LinkedOut?

How many invites to refer one member of your network to another have you received? Isn't that the real measure of the value of a social network site, e.g. whether it enables connections to be made that would otherwise be impossible?

My network reflects the relationships I already have. If it doesn't enable something more, it's just a glorified contact book (admittedly one in which I get an interesting glimpse of my friends/colleagues job aspirations trough their assertions as to who they are interested in hearing from :-)).

I like LinkedIn and I enjoy the feeling of nurturing my set of connections through invitations to join, but I can't say I've benefited from the supposed 'network effect'.

Thursday, September 22, 2005

It's a Flat Flat World

Reading Thomas Friedman's The World is Flat- centered on how the lowering of trade, political, and technological barriers now allow people to do business with others located all over the world.

For me, one of the most interesting implications of this gloabl connectivity is how the time zone difference between the different regions can be leveraged and taken advantage for increased productivity and efficiency. Friedman cites US hospitals that, overnight, send X-rays to India where they can be read by trained radiologists during normal (Indian) working hours.

I've experienced this sort of time-shifting first hand. My NTT colleague Yuzo Koga and I are co-editors of a specification within the Liberty Alliance's Web Services Framework. The fact that Koga-san and I are on opposite sides/ends of the world (he in Tokyo, myself in Ottawa) proves a blessing whenever we are faced with an editing crunch. When one of us finishes for the day, he simply sends the latest version of the spec to the other, who is just beginning their work day. There is little dead time where the spec is not being actively worked on. The 13 hours between us allow us to work "together" far closer than we ever could in the same city.

It does seem ironic that the book's chosen metaphor for the new geography is one that would actually make impossible this phenomena - if the world really were flat, then there would be no time zones to shift work through.

Friday, September 16, 2005

Tag Clouds (or Cloudy Tags)

Seeingthe new home page of Flock got me thinking about whether anything interesting could be done with tag clouds. I wonder if anybody has tried to game the weighting Algorythms to artistic purposes.

Tuesday, September 13, 2005

Descent by Modification

If I was starting over and could choose a different career path, it would be evolutionary genetics. I guess that's why I find myself constantly looking for related analogies within my current career. And so that's why I found the matrix of SAML 2.0 conformant products (as determined by the Liberty Alliance) so interesting, specifically the feature support of HP and Trustgenix.

The rows for HP's Select Federation products and TrustGenix's IdentityBridge product are identical in the SAML 2.0 feature's they support. Given the number of permutations of possible features (even after accounting for the logical dependencies that would decrease the full number), this perfect overlap suggests a relationship between them, just as the large overlap between the DNA of humans and (for instance) chimpanzees indicates an evolutionary relationship between them.

That's why I was so pleased to track down this press release describing just such a relationship - specifically an OEM one. The two products share the same SAML 2.0 feature support because, in a sense, HP's product is descended from that of TrustGenix (at least with respect to support for SAML 2.0).

I'm not suggesting that we are descended from chimps. maybe that's a shame - pygmy chimps (or Bonobos) do seem have to found a gratifying lifestyle.

Monday, September 05, 2005

Universal Identity Grammar

Reading Steven Pinker's Blank Slate again and revisited the concept of Universal Grammar (first proposed by Noam Chomsky)- the posited common underlying rules/infrastructure on which all the world's various languages are built. The theory is that we all have an innate instinct (a language instinct) for these rules and that, based on which specific language we are exposed to as children, various switches are thrown (determined by the environment in which we are raised) that determine the actual manifestation of those rules (e.g. which language we speak as adults). Its asserted that that is why learning a language as an adult is so much more difficult that as a child, you're fighting against the switches that were set long ago.

Universal Grammar suggests that the underlying structures of language, the grammar, is innate and the same for all humans; different languages are the result of some config file in the head of a young child being set with various binary parameters. The simplest illustration of one of these params is the choice of Head first or Head last; depending on which choice is made, a language is either SOV (Subject Object Verb) or SVO (Subject Verb Object) with many associated orderings in other aspects of syntax. For instance, Japanese is a SOV language and English is typically SVO. In Japanese, the verb always appears at the end of clauses and sentences (a fact which makes my sometimes attempts to learn the rudiments of Nihongo interesting).

Given the recently introduced concept of an identity 'metaystem', the idea of some set of fundamental set of rules and/or components, from which various identity applications can build, seems appropo. Just as languages differ in their particular manifestation of the Universal Grammer, particular identity systems will inevitably vary in their manifestation of the basic underlying components and principles.

I can imagine that in the 'mind' of every young SSO architecture, there is a config file with switches for 'front-channel vs back-channel', 'remote or local storage', '3rd party or self assertions', etc. The specifics of the environment in which that architecture (e.g. B2B, constrained client, privacy requirements, legislative issues) grows to maturity determine how the various parameters are set, and consequently the use cases for which that architecture is appropriate as an 'adult'.

So I guess we shouldn't feel bad about favouring one federated identity architecture over another - it's just that we are adults and our switches were set long ago.