Archive for the ‘Tagging’ Category

How many ways to perform “inside search”?

April 20, 2007

Search imageI’m fascinated by the many different approaches to search that are emerging to challenge the “inside search” we all know and love (hate) on web sites. You know, where you type in a keyword and get a large and random set of search results returned from a basic keyword index.

I’ve written about Ultra Knowledge before and how their automatic tagging sets them apart from the crowd but now the crowd are starting to take differing (and potentially complementary) approaches.

Collarity sits on top of the existing site search and monitors where the users click through to for each keyword. This then creates a “wisdom of the crowds” intelligence so that when the next user types that keyword, the system will suggest the most popular destination URL’s. This is interesting as Collarity don’t actually index the content which makes it fast to deploy. The downside of this is that it’s only as good as the underlying site search so if that doesn’t return relavent content then the value of the wisdom is reduced to the lowest common denominator.

Eurekster takes a different approach, allowing the users to rate the search results thus increasing the weighting of the most valuable results. It also allows users to create their own search results based on their knowledge of the subject making it a mashup of search and wiki functionality.

And this week I met with a UK company called Synature who have yet another take on “inside search”. Much more focussed on specific sectors (initially holidays) their product is about connecting like minded users using something they call “Attitudinal Matching”. They employ a psychologist who sets a series of 4-6 questions about how you feel about a particular subject. For example in holidays you would be asked to rank your requirements on a scale between “Beautiful” and “Exciting”. The results of these questions are then fed into a sophisticated system that can match you to like minded people and their holiday recommendations. Again, the system doesn’t index the content, just the user who created it and allows you to locate relevant content via the social network that this creates.

There are many others of course, if you know of anything interesting going on in the search space, please leave me a comment as it’s such a fast moving area it’s impossible to keep up.

Which is right? Well all of them have their pros and cons in different applications and the possibilities for combining them with each other and the social networking platforms is intriguing. Watch this space…

Posted in Search, Social Media, Tagging, Web 2.0 | Leave a Comment »

User generated content, what are they talking about?

March 23, 2007

One of the major components of the Web 2.0 revolution is User Generated Content or UGC to its friends. But while it undoubtedly adds colour and content to the traditional editorial on a site, just how useful is it?

The main problem with UGC is that nobody knows what the content is about. It’s difficult enough getting editorial teams to classify content consistently, so asking users to do this is unlikely to work. Of course we’ve seen the concept of tagging work brilliantly on the likes of Flickr and del.icio.us but even here there are a lot of inconsistencies. People don’t all use the same language, have the same idea about what the content is about and, some just can’t spell.

The upshot of this is twofold. Firstly, its pretty difficult to automatically connect editorial content about a subject to its UGC cousin. There’s no consistent tagging or mapping to enable this to happen automatically. So for example if I were reading an article about tagging, it would be great to see all the member blog posts that were related to that subject.

The second is more important to publishers which is that there is no easy way to target advertising at the readers of UGC. So despite paying $580M for MySpace, NewsCorp couldn’t run any of their conventional display advertising on the site. There was no targeting available other than the basic site demographic. They have solved their commercial problem in the short term by doing a $900M deal with Google to run AdWords on the site. Big advertisers though who want to reach the MySpace audience don’t want to do this through AdWords.

They want their brands and imagery associated with quality, targeted content, not mixed in with all the random ads in AdSense for unrelated products. If you have any doubt about this, have a look at the ads that get served against your emails in GMail for seemingly unrelated products. It’s a great game and depending on the subject matter you can expect to get some spectacular results including one example I found that combined British Airways, satellite TV, the best tennis grip and lingerie. I’ll leave it to you to work out what the email was about…

So what is the answer? Lots of sites have introduced tags as navigation devices either in clouds such as or associated to the stories as on my fovourite site ZDNet. The holy grail is to automatically tag all content consistently across the site whether its editorial, UGC or vendor content such as whitepapers. That creates a consistent way to connect content together and also to identify what its about so as to be able to target advertising.

The new search engine on ZDNet, provided by UltraKnowledge is designed to tag all content consistently regardless of where it originated from. At the moment it is only indexing the editorial content but soon it will be rolled out to the vendor and UGC areas. At that point, it should start to provide the targeting the publishers require to make money from inventory which was previously low value, run of site.

Posted in Ad serving, Social Media, Tagging, ZDNet | 2 Comments »

Searching without thinking

November 27, 2006

For too long search design has been dictated either by the search engine technology i.e. “Keywords” or by the agregators (Google/Yahoo/MSN) ranking algorithms. Both have their benefits but neither necessarily serves the user the best.

Keywords work perfectly if you are searching for a product. If I want to know about the Treo 750W then typing that into any reviews site or search engine I’ll get very specific results. However if I wanted to find out about Smartphones, then things start to get messy. Firstly the term Smartphone means different things to different people. Secondly most product reviews probably won’t even mention the word “Smartphone” in them anyway. So keyword searching on a topic is a very hit and miss affair.

Using the main search engine sites, the problems are different but equally vexing for the user. At the mercy of the algorithm, users are now provided with content that has been deemed by some invisible mechanism important.

Google changed the landscape here by working out that what the users thought was important (i.e. Number of links to that content) was actually more relevant than the most recent content with that keyword. Of course, it’s not as simple as that and the big three spend millions of dollars tweaking and tuning their “special formula” with as much secrecy and security as Coca Cola.

Natural language search engines have been around for a long time but the problem has always been that it takes around 100 times more computing power to index and search in this way compared with conventional keyword search. Now, with processing power becoming cheaper and cheaper, we are starting to see investment ramp up in these technologies. Cited in this article, there were 47 search start-ups in 2005 raising a combined $260M.

And here on ZDNet, we are starting to experiment with this technology too. Our new search engine has the ability to automatically create relevant tags and topics related to the search term. It’s still early days for this type of technology but we believe that this type of search is the only one that considers the user first. Up to 75% of search terms are a single word and our most popular search term is often “Microsoft”.

Given the size and scope of Microsoft’s market, this would lead you to believe that the user doesn’t know how to search for what they are really looking for. And the truth is, most users are working blindfold when it comes to understanding conventional search syntax so they go in one direction until they bump into something, re-adjust and try again.

By creating tag clouds of the most relevant tags and terms, we are narrowing down the search for the user and helping them towards what they were really looking for. Go on, give it a try, take off your blindfold, click here and check out the related tags box.

Posted in Search, Tagging, ZDNet | 1 Comment »