Speakers from Microsoft, Blinkx and Last.fm discussed issues of content regarding search, recommendation, the semantic web and the ownership of data in the Web 2.0 era at Content 2.0 on 6th June 2006...

Report by Deirdre Molloy

[Register and post your own comments on this article below...]

Download this session from the Content 2.0 Podcasts!

Chair Mike Grehan recounted that there was time a time when he manipulated search results so much for his clients that his name was on the wanted list of every search engine. Page Rank, he explained, was the brainchild of John Kleinburg who developed the idea of "hubs and authorities". But Page Rank’s credibility is overrated - linkage data is just about the abilities of the guys who build and manipulate widgets, he observed.

Suranga Chandratillake - Blinkx

Suranga explained that Blinkx searches video content – of both the garage and broadcast-generated varieties, and its Web 2.0 features include RSS and embedding. Blinkx looked at the video data first of all, and now speech recognition, and this has enabled them to build a very fast scalable database. Their Pico service is still in (public) stealth mode.

In terms of the history of search, Suranga was very much in concurrence with Bradley. Search as a user-focused service is great when you’re in that active mode but not so great when you are the one who is passively consuming. The photograph of the baby curled up with the pet on Flickr was great, but we use the links – we don’t go and search Google for babies with pets. Pico tries to take this to the next level – it’s a bit like interestingness.

Search is so little solved, he commented, that they are doing fine in their niche. Blinkx TV just has all the same problems that the GYM (Google Yahoo! Microsoft) search engines have, but they are looking more at how video seems to be about the topic. In this regard, he described Pico as a contextual tool as its technology looks at the entire page text.

Alex Barnett – Microsoft

Alex flagged up the topic of attention and what he means by it. A study at University of California Berkeley in 1999 and 2003 looked at how much information there is in the world. In 1999 there was 2-3 exobytes, and 92% of it was on magnetic tapes. By 2003 there was 5 exobytes. Only a fraction of this was on the online “shallow web” – just 0.003%. In the deep web there was 97 terrabytes.

In ‘The Attention Economy’ Michael Goldchaber described how information was becoming commoditized, and Steve Gillmor has developed this with his attention.xml. So, Alex continued, how do we design software and experiences so that people can find this content?

"We need to ensure that all those attention data types are available for the customer to own. Then services can be created to plug into it"
- Alex Barnett, Microsoft

He also raised the notion of “my data”, and argued that content needs to have APIs so that 3rd parties can create value from it. So what is attention data? It includes wishlists, blogrolls, photo collections, buddy lists, favourites, and your search history. All these things are my personal history, he stressed, and my history (or profile) is mine – I’m generating all this attention data and we need to ensure that all those attention data types are available for the customer to own. Then services can be created to plug into it. I’m then able to use these “attention filters” as filters into the mass of data.

Mike Grehan interposed that Google is still trying to play catch-up with Yahoo! in gathering data on users relying on log-in. But, Alex responded, why won’t Amazon allow you to take your wishlists elsewhere? As Yahoo! has shown with Flickr, there is more value, credibility and trust to be leveraged from allowing people to export their content.

Matt Ogle – Senior Web Architect, Last.fm

People take different approaches to using Last.fm Matt explained – you can just listen to your stuff by downloading the (Audioscrobbler) Last FM client which he termed as “myware” – it talks to your media player and gathers data: what you listen to, when, etc... and they now have 2 billion distinct pieces of song data.

Collaborative filters are another feature of the site, and Matt observed that there are also short-term groups of fans on the site that form and dissolve. He spoke of their Long Tail slider – “obscuromatic” – and explained that they built customised radio stations for users on the back of all this data. They also offer blogs, music journals and other Web 2.0 filters all of which are filtered through into the recommendation engine. Entries are also filtered by musical tastes.

"A lot of our new services are about friends and contacts, and you can restrict activity to your friends list"
- Matt Ogle, Last.fm

You do give up a certain amount of privacy in return for these services, he acknowledged, but Last.fm want to open up their RSS feeds so you can get an API to do an entire data dump. On the human side, a lot of their new services are about friends and contacts, and you can restrict activity to your friends list.

A delegate asked if he thought the music recommendation service Pandora – in contrast to Last FM’s community and social approach – pandered to the selfishness of consumers? Matt responded that he agreed that Pandora has nailed personalised radio.

"Discoverability is a huge thing that attention data brings to the value chain, and it’s people that define the content quality, not an algorithm"
- Alex Barnett, Microsoft

When will machines start discovering stuff for me, Sam Sethi asked, so the machine feeds me rather than vice versa? Alex Barnett interjected that Delicious software enables the consumer to discover the work of another. This kind of discoverability is a huge thing that attention data brings to the value chain, and it’s people that define the content quality, not an algorithm.

Suranga explained that Pico (unlike Delicious) requires no log-in – it sits on your computer and if it thinks it has found something relevant to what you’re doing now it will alert you. Matt added that through RSS, Last FM offer processed versions of your attention data – although computationally this is very expensive. They are also moving recommendations onto the Last FM Dashboard which is only viewable to you, a bit like your rolling news, he commented.

Mike Grehan made the interesting point that allowing yourselves as a company to lose that data, as Last FM do because they allow people to remove their data, seemed paradoxical to building an intelligent system where you get people to describe their preferences.

Sheila Sang of Interactive Powwow wondered if, from a content development approach, there was anything the panellists were working on and Suranga flagged up they can insert a video search function into your website.

"The idea of something monitoring what you do on your computer is still a strong and valid one"
- Suranga Chandratillake, blinkx

Miko Coffey of NESTA commented that the description of Pico reminded her of the Microsoft Word Paperclip and asked were Blinkx walking a tightrope between interruption and giving people what they want? Suranga addressed this saying that with the paperclip story, the idea was good but the execution was not. The idea of something monitoring what you do on your computer is still a strong and valid one; the issues come with how well it’s done. Pico has to be subtle and you can turn it off. Privacy was a big concern for his team prior to launch, but they have only had two bloggers commenting on it. If you’re clear about what you are doing with the information then you can guarantee security better with a system that isn’t entirely open, he said.

Matt observed that Adsense and other banner ad companies have allowed them to grow, and they can also sell and discover audiences for brands. Alex Barnett distinguished between free ad-spammed services and premium ad-free services. If information is the value, he reasoned, how do I differentiate my recommendation data and cross-reference my preferences? For example, how do I filter my music preferences based on my book preferences?

"We are still living in the library model of the internet, but we should be allowed to swim through that data"
- Suranga Chandratillake, blinkx

Suranga said that access across multiple PCs is easy; the trick is how to get the information without it being complicated and messy. Alex outlined the scenario whereby you can pull out your OPML file and drop it into your YouTube or Technorati page – because silo-fication is not reflective of real life. Capturing the terms for the semantic web is nearly impossible, as Tim Berners-Lee has discovered, he added.

Nico Macdonald of Spy and London AIGA Design pointed out that visualisation of the web is not used, but this also relates to how we represent ideas and people through search. Alex picked up on this recalling how in the early 90’s the web design standard of left-hand navigation was cemented, then you had the Minority Report-style sites but in usability testing it turned out that people just want simplicity and familiarity. In search results there are multiple user interfaces provisioning connection back into the data, so the market is voting with its feet, he reckoned.

Alex also admitted that whether it’s RSS or OPML – people don’t care. Technology is just an enabler but, he predicted, the younger generation will be aware that it’s their data. Suranga reflected that today we are still living in the library model of the internet, but we should be allowed to swim through that data, and it doesn’t matter how you find that data. For now, he concluded, we’re still on the outside peering in.

Content 2.0 - 2006 conference Website:
http://www.content2point0.com/2006/

About Suranga Chandratillake:
Suranga Chandratillake, co-founder and CTO of blinkx, has lived and breathed tech since he was seven years old, when he was first introduced to the 6502-powered BBC Micro Model B. He went on to complete an MA in Computer Science at the University of Cambridge, England, where he specialized in distributed processing architectures. Professionally, Suranga has worked on a diverse (the unkind might say confused) range of projects in a variety of roles—as an algorithms developer on portfolio risk balancing systems for the investment bank, Morgan Stanley Dean Witter, a technology analyst on voice-recognition applications at netdecisions (a UK-based consulting and venture technology group) and, catching the start-up bug, founding Anondesign, a Cambridge-based technology company that built a ‘live’ content management system years before RSS was a glint in someone else’s eye. Finally, another Cambridge-based technology company — Autonomy Corporation PLC — caught his attention and he joined the company. After working through a number of roles he moved to San Francisco to be the company’s US CTO - the role he played directly before founding blinkx.

About Alex Barnett:
Community Program Manager at Microsoft Corp, Alex has been creating innovative, useable and successful online customer experiences since 1994 and has a deep passion for the web - for its limitless potential to improve people's lives and to create new businesses opportunities. He has pursued this passion with some of the greatest companies in the world (examples include - Microsoft, Levis, Reebok, Volkswagen and GE). Originally from London, Alex joined Microsoft UK in 2002 as Online Customer Experience Manager (CRM and Online Marketing). Previous to that he was a director at Bluewave for 7 years (an international online communications agency). And before that he was doing something completely different – he was a professional cricketer for 8 years (Middlesex & Lancashire). He blogs at http://blogs.msdn.com/alexbarn/default.aspx

About Matt Ogle:
Matt is a Canadian web developer who was worked in London since 2004. He is currently busy fomenting the social music revolution over at Last.fm. Matt is the creator of the Streetprint Engine, an open-source CMS for sharing and archiving printed ephemera, for which he received his MA in English Literature and Computing Science from the University of Alberta (Canada) in 2004. Since joining Last.fm in early 2005, Matt has helped guide the merger of Audioscrobber.com and Last.fm, and shaped many of the site's user-facing recommendation, communication, and music discovery features. He plays keyboards in the company band.

About Mike Grehan:
Mike is founder and CEO of MarketSmart Interactive Ltd., a wholly-owned European division of MarketSmart Interactive Inc., the world's largest search engine marketing company. He's recognized as one of the foremost SEM experts and is the author of multiple books and white papers on the topic. His best-selling second edition of Search Engine Marketing: The essential best practice guide, received more plaudits than any other on the subject from the industry's leading players. In 2004, Mike is a sought-after search engine marketing speaker, and his own newsletter has attracted over 17,000 subscribers. Mike played a key role as founding member and promoter of the global Search Marketing Association (SMA) movement and sits on the board of SMA-UK. He blogs intermittently at Mike Grehan Says.

OTHER CONTENT 2.0 SESSION REPORTS

Content 2.0: Mesh Up - Connecting Content To People

Content 2.0: Goodbye New Media Hello Social Media

Content 2.0: Marketing 2.0 Forum

Content 2.0: Can Brands Be Trusted?

Content 2.0: The Future Of Web Search

Content 2.0: Folksonomies - What Are They Good For?

Content 2.0: The Invisible Culture

Beers & Innovation (music special) @ Content 2.0