Latest Interviews
‘The Search Goes On’ – Opinion By Michael Nutley.
Posted: Monday 2nd Apr 2007
Search has been the accepted way of navigating the internet for so long that it’s sometimes hard to remember how things were before; before Google became a verb, before search marketing became the biggest part of companies’ online advertising spend, before anyone had thought that organising all the information on the planet might be an achievable goal.
But even though search has become a staple of our online existence, for the past few years its development has seemed almost static. The search engines themselves have introduced all sorts of innovations, but the core of search seems not to have moved on very much. The discussions about the limitations of search still focus on the same problems that they did four years ago; that most searches don’t give you what you’re looking for, that no-one needs a million results from a search query, or even looks beyond the first couple of pages.
At least part of this is our own fault. We’re actually very bad at using search, choosing to type in very short query strings and complaining when the engine doesn’t deliver what we want. But without a breakthrough of the type that established Google’s current dominance in the market, people are starting to talk about what comes next. Could we be entering the post-search era of the internet?
The competing approach that is causing excitement is user-generated content in the form of recommendation and tagging. Algorithmic search, the argument goes, is great for finding all the locksmiths in Camden, but very poor at identifying the best locksmith in Camden. For that you need a different approach; recommendation.
Recommendation and tagging are key parts of the Web 2.0 toolkit, typified by Amazon’s use of user-generated reviews to help persuade customers to add further purchases to their shopping cart. Subsequently, review sites have proliferated, followed by a growing wave of sites based on the “people who bought this also bought this” approach also pioneered by Amazon. Online music recommendation site Last.FM is a perfect example. But there’s another type of site that has also become astonishingly popular, the type that allows people to answer each other’s questions and rewards those providing the best answers with increased status within the community. Yahoo! Answers, which has been hugely successful in Korea since it launched there last year, is a great example of this approach.
User-generated approaches such as these suffer from a number of problems. There’s what experts call “the bootstrap problem”, where getting enough content to make the recommendation service work relies on the altruism of a few dedicated taggers and recommenders. Interestingly, the increased status that comes with being a recognised authority seems to go a long way to addressing this. Recommendations can also be unreliable, either for amusement, as anyone who’s read the user reviews of David Hasselhoff’s albums on Amazon can attest, or for more sinister reasons. While it’s easy to tell someone who’s getting negative reviews that what they need to do is fix what they’re doing wrong, it’s much harder to stop competitors posting damaging comments.
But while a lot of companies, notably Yahoo!, have spent a great deal of money on tagging-based approaches, buying Flickr and del.icio.us among others, another approach is starting to emerge. The possibility of combining natural language recognition with artificial intelligence is giving rise to semantic search, which would allow users to type the sorts of questions they currently ask on “answers”-type sites, but have them answered by machines. This would have the advantages of avoiding the bootstrap problem and being harder to game than recommendation services. It would also be far easier to scale, and while it wouldn’t replace recommendations of the type employed by Amazon, it would give computer-based search h a much greater utility. And, like previous innovations before it, it would create a whole new industry around making sure the machines collected the information you wanted in creating their answers.
Michael Nutley is editor-in-chief of New Media Age.