|the next generation se's|
Today we have about ten or so directories and engines that dominate the search engine world. surely these engines will be replaced by more advanced methods of searching any ideas what they will be?. there have been lame attempts at graphical directories surely there has to be an advance and soon. what do you all think will be the backbone to the new generation of se's
[businessweek.com...] is an interesting article about that.
The next generation of search engines will be:
1) Faster and more complete. Hard drive space should be down to a dollar a gigabyte by q2-2002. Maxtor just released a 100gig drive for $294 in quantity. IBM has announced they can squeeze 480gig on to a 3.5 standard sized drive. The easing of the storage space issue will allow se's to store more data than ever before.
2) Speed. Prices on ram are in a downward spiral while motherboards are supporting vastly higher ram densities. The fully populated 1 terrabyte motherboard will be in reach by the end of 2002. That will allow se's to store their entire primary key file (main index into the database) within the memory of each work station. Also, there has been talk of another major breakthrough in chip sizes that could allow as high as 20mhz machines by as early as 2004.
3) Interface. Despite claims of cable and dsl broadband ruling the world, se's have to pitch to the lowest common denominator - 33k-56k. It will be that way for some time to come. I don't expect anything radical in interface changes in the near term future of se's.
4) Rounding out the features. Now that most se's have their system primed and the major bugs squashed, they now have "play time". That extra time can be spent flushing out features like topic specific searches. Googles UncleSam and EDU topic niches are brilliant and I expect to see more of it. NorthernLights Special Collection is also good.
5) Pay-for-Play backlash. I think we are already seeing it. Some engines that have switched to PPC services are reporting significant drops in traffic. People want to know if they are looking at the YellowPages (paid ads) or an Encyclopedia.
6) The subscription SE. I think any one of the top se's could move to a subscription based model and make a go of it. One of them is going to try it sooner or later. Leave the ad format in public, while the users behind the scenes get the full monte.
I'll bite, what's splunge?
The UK version of "punt". aka: fresh out of ideas. (splunge comes from another montey python skit. A guy couldn't think of anything to say when put on the spot, so he said "spluge" and everyone repeated it as the right answer.)
Brett hit the nail on the head with his list. I've been trying for weeks to think of something to add to the list of what the search engines of the future might hold, and have come up with nothing.
Subscription model I think might happen. Any takers on when?
Database size is always increasing. Over the last year and a half, between google and alltheweb, we've seen the size that a search engine can hold in terms of data grow incredibly.
Speed of results has gotten so fast, in some ways I can't believe it. How fast will those be once net connections rise, even for the "lowest common denominator"? Even faster still.
Relevance: we are seeing new definitions of contextualized search relevance all the time. Look at the algorithmic rotation done by google, it is phenominal to witness a refresh of an entire database so large, so diverse, across new ordering schemes on a regular basis. Who knows but that might be the key to the next generations success.
As far as information, I think that more and more SE's are beginning to realize that some human element integrated into the mix is a good thing: Google's acquisistion of Deja's usenet groups archive could be seen in this light. Perhaps they are using that as a basis for algorithmic benchmarking for natural language processing? If they are, they might have newly relevant ways to parse data in the next few years.
Features, relevance, speed, and the ability to actually give you what you are searching for is the future I see for the engines.
I'm not sure that this will be in the 'next' generation, but I would suggest auto-translation will start appearing, as in 'click here if you want all results presented in English as default'. That would be invaluable for European searches.
>>2) Speed. Prices on ram are in a downward spiral while motherboards are supporting vastly higher ram densities. The fully populated 1 terrabyte motherboard will be in reach by the end of 2002. That will allow se's to store their entire primary key file (main index into the database) within the memory of each work station. Also, there has been talk of another major breakthrough in chip sizes that could allow as high as 20mhz machines by as early as 2004.
Im thinking you mean 1 terrabyte of hard drive space. RAM is still a little short of that. Few SERVERS have more than 2gigs RAM. Also, 20ghz processors I assume? I'd also consider stuff like parallel systems...look at IBM's Blue Gene project and Linux Beowulf Clusters.
Interface WILL change. While many of the internet COMPUTER users will still need the same old interface, there will also be palm users, WAP users, and even BMW is thinking about putting google in their cars.
User tracking directhit style will become more prevalent, although it already is. I also see onsite variables losing more and more weight.
I see the possibility of sites like amazon paying out large dollars for a subscription type service that dynamically places their specific product links directly into the search results, integrated directly with their product databases.
SE's like google may even start their own ecommerce shops and sell products simply through the searches. Google definitely has the consumer trust and user profiles to do this well.
Lots of changes still in store.
A possible long term future,
An Open source distributive database that is based on page analysis as well as surfing habit analysis. The page data would be collected as the surfer views the page.
The algorithm would be some type of mix between kw density analysis, site theme, and surf habit relevancy projection. There would be zero reliance on link popularity.
"Based on previous searches we believe you may like the following sites..."
"You viewed x site, this site has updated in the last week" (now that would be cool)
Privacy, smivacy- give me features.