The Future Of Search? It’s 4D Isn’t It?
People in our industry are often asked what they think the next developments will be in the growth of the internet, and in particular, search. Well, using the example of how frustrating a recent car journey was, I think I can probably make a pretty good guess.
Historically search has been two-dimensional; what I mean by this is that when a query was made the only thing that the search engines ‘knew’ for sure were the word or phrase that had been entered, and the time of the search. But with a variety of developments in technology, it strikes me that search engines should soon be able to move into 4 dimensions.
The first of these new dimensions is that the search engines can now call on knowledge of a user’s likes & dislikes. Google already compiles a record of a user’s web history, which tracks not just what results they have clicked on but all the sites they have visited; whilst it can only do this with people who have Google accounts of some sort or another, with the growth of Gmail, Google Analytics, Blogger and other services, this is a fast growing number.
MSN & Yahoo! also have access to this sort of data through their own services such as Hotmail. What they, and Google, also have is increasing amounts of data which allows them to tie people into what is known as the social graph. By buying sites such as delicious, Flickr, YouTube & the like, the engines have access to huge amounts of data in the forms of tags. They can then tie this in with the same data of ‘friends’ and peers of their users, in what many are referring to as the ‘social graph‘.
Finally, search engines are increasingly making more & more use of geographical data. As the advertising for the Nokia N95 phone showed, the engines and the mobile networks are keen to tie in search with the GPS data that can be retrieved from a GPS phone. This should mean that when someone makes a search from their phone they can be provided with geographically relevant data on an incredible scale. Whilst engines have been using IP data for quite some time, this is a major leap forward.
So, how does this all tie in with the car journey I mentioned earlier? Let me explain.
I was driving when I decided that I wanted a coffee. Not being familiar with the area I needed to find out where the nearest Starbucks was (I’m very loyal to the lady of Seattle). Now I have a car with voice recognition; I have GPS on my sat-nav; and I have web access through my Blackberry. And with all of this it seemed obvious to me that I should simply be able to ‘ask’ my car where I could get a coffee.
The voice recognition would input this into the search engine that would know where I was and, via my web history & social graph, that I prefer Starbucks to other brands. The results of this search would then be fed into my sat-nav & I’d be minutes away from my Half Fat Soya Free Frappawackachino.
Whilst this may all sound far-fetched, when you consider the fact that BMW & Google recently announced a tie-up, and look at the sort of search results you get from Google maps, it’s really just a matter of tying all of these factors together. And whilst I wait for this to happen, and fume at the fact that something that seems so obvious, and is (in theory) perfectly possible, doesn’t yet exist, you might want to look at Ciaran’s recent guide to optimising for universal search, as all of those factors are only likely to become more & more important.
Another thing Google know: exactly how much profit (or loss) you make for them (since they can track your ad click-throughs).
I actually don’t like IP/country-specific search results–especially since there’s no indication that if you did the search in another country, you might get quite different results. (Microsoft’s various properties keep trying to sell me Australian holidays and Australian mobile phone contracts, three years after I stopped living there.)
Compare:
http://www.google.com/search?q=grand+national
http://www.google.co.uk/search?q=grand+national
http://www.google.com.au/search?q=grand+national
Maybe they need sliders that indicate and let you control how tailored the results are to your location, previous search history, etc.
Hi Dave,
Good thing to point at universal search at the end of the article. Geo-localization (along with mapping) is a hot topic, geo-marketing strategies have been thought, but we’re just starting to see some geo-centric Webplatforms: Fat Door, Peuplade. Still shy to me.
Hi Dave,
Cracking article. I agree 4-D is not far behind though as an ex-pat I find that sometimes geographic/IP settings can be a hinderance if I am planning a trip home or trying to find a nice place to get a gift certificate for friends living abroad.
I also think no one has successfully cracked the UK geo-location thing yet as there are so many different ways to slice that pie (telly regions, county, country, post code, north, south, etc) which makes it more difficult than when dealing with the US. Who ever can prove a leader in the geo-localisation of the UK will be laughing all the way to the bank.
Cheers,
D
Dave,
you touched on some really interest points, mainly the fact that with today’s technology we have access to a lot more informational dimensions associated with a search than in the days when what you have dubbed ‘2-D search’ was incarnated;
And also the shortcomings of today’s search engines that apply a one-size-fits-all relevancy model that they bucket all users into.
Fast Search & Transfer - the leader in enterprise search, have identified these shortcomings and developed a unique relevancy model to both capatalise on all these new attributes of information available to us and put the power back in the hands of the user - allowing her to tune each of the relevancy attributes to their personal preference.
Google bases their PageRank algorithm on ‘link cardinality’. That is, the more links to your site the higher your rank. Furthermore, the more high ranked sites that link to a site the higher the rank.
A clever approach but some other dimensions we can use are
# How fresh a document is? An article published on Iraq now versus in 2003.
# How well does the query match superior contexts like the title or the URL?
# How statistically relevant is the document to the query - word proximity, word frequency, word similarity,.. ?
# What is the quality of the document? Press release vs. a blog entry.
# What information can be extracted from the users profile?
# In which parts of the document do the query terms appear - URL, title, abstract, body, meta-data,..?
# Geographical distance- If I search for pizza parlours from London getting results from Milan is not to helpful!
#…
In turn these are sliding scales that we can adjust according to our preference. Like the scales on a graphic equalizer.
In 2006 the IDC estimated that we created, recorded or copied 161 exabytes – or 161 billion gigabytes. In volume, this is equivalent to the amount of information in all the books ever written – multiplied by three million.
This provides a huge challenge (many say opportunity) for search to harness this information in a way that is meaningful and relevant to the user.
K
You’re right that there is still lots more to come, but if I read your comment right, you’re suggesting that all the things in your list are still on the horizon:
In fact, as far as I’m aware these are nearly all factored into the ranking algo nowadays except for the profile stuff, or at least not to the extent to which it could be, and that’s really where the changes will be.
Dave,
A really interesting article and you touch on some really interesting points. I’d be interested to hear your views on the current geo-political environment and how future interest rate hikes will affect the consumers interest in the Internet