Raul Valdes-Perez of Vivisimo begs to differ with all the hype around personalized search (including in my book), and the idea of major engines mining your clickstream to better understand your intent (and give you more personalized ads, of course). In a short paper outlining his views, (PDF download), he outlines five major problems with personalized search and concludes:
…. search personalization is likely to waste the talents of top computer
scientists. It may even give worse results…
I totally agree with everything Raul says. Best line is the last: “The best personalization is done by the persons themselves.” We’ve built our entire technology/company around this belief.
V. smart analysis of why personalization is a mistake.
The direction is not to personalize – it’s to relieve disambiguity. The problem with search is that we use the same word in multiple ways. The solution should be simple, ask the user which definition they meant by the term entered. That’s what wikipedia does. The future of search is enabling smarter searchers, not creating a psychic search engine.
On the surface Raul is correct however search visionaries are thinking floors above the surface.
One day personal search results will be refined to exacting results.
Somewhere at the end of a dark hallway at IBM, Google or FAST are groups working on just that.
Personalization can produce better search results. I recall reading an CACM article on personalization from some folks at PARC that showed encouraging results, despite the challenges posed by Raul.
Just because it is a difficult problem, does not mean it is not worth pursuing. In fact, just the opposite is true – the harder the problem the more value is created in the market by solving it.
For the casual surfer the advanced Search Features may be too esoteric, they have just began replacing one and two keyword queries with three and four keywords, many do not even use the local search (they just add the city or state in the keyphrase query)
The Yahoo Mindset is a good compromise between General Personalize search and the specific personalized search.
If someone is doing an ongoing project on Google – it may be efficient for them to have their search proclivities default to a certain theme.
However, since personalized search is often an opt-in feature, it does allow those who want to turn it on or off to do so.
Click tracking can also be an asset when calculating the proclivities of a searcher.
But like everything, we are at the dawn of this techology – it will evolve and refine itself in the upcoming years.
Great article. I also prefer the approach of adding search criteria explicitly rather than implicitly.
I would much rather deal with an engine like Vivisimo that at least attempted a conversation with me but sometimes made strange suggestions than deal with a “black-box” engine which silently records what it interprets as my preferences and makes little or no attempt to articulate them.
HE says correct.
But think is marketing View point. Narioki kanus. Delighters way. Here after people doesnt use or beneffit from the services every once in a while google has to role out some thing. To delight people.
Why cherry coke , Vanilla coke and all. Do you know how many percentage of people drink that. Amazingly low.But just to make people think that they are deligheters . I mean some thing new everytime. We are going to see more of this any way ..
VJ
Another excellent paper that Raul wrote a few years ago is titled, “Needed: A More Selective Ignorance.”
http://www.vivisimo.com/docs/overlook.pdf
Bottom Line:
“A discriminating ignorance is bliss.”
I would also agree with the comment about Yahoo Mindset. The technology and concept is perhaps better seen with Yahoo’s Smart Sort.
http://shopping.yahoo.com/smartsort/
I enjoyed reading the article and agree with Raul. Perhaps we don’t need personalization. Rather, lets focus on the masses. I think the value add comes in to play when you can impact my results based on what others using the same and like search terms have been clicking through. No, I don’t want my results manipulated by what the masses are doing, but lets face it the study of sociology indicates that human behavior is easy to predict when you look at the masses. The reason for this is that we in fact are not that much different from those around us. We think we are, sometimes we really think we are, but we aren’t. Where am I going with this? 999 out of 1000 users a likely the same or very similar to the point that a massive database of search activity or ability to track searches beyond the simple query would likely yield relevant results based on what others have searched for and clicked on. Goes back to John’s idea of ‘The Database of Intentions.’ The concept that I think many are afraid to come to terms with is the idea that if I am searching for something, someone has likely already done the same.
I disagree with pretty much everything he says. None of the problems he mentions are particularly cumbersome.
It really comes down to expectations. I don’t think personalised search is going to make search engines right anywhere near 100% of the time, however the incremental improvements will be worthwhile.
Let’s people personalize the ways they are personalized.
Personalisation?
5 words:
“My TiVo thinks I’m gay”