In Ch. 6, Global Swarming, of Andy Clark's book he mentions "collaborative filtering." The average person might not know what this term means, but I'm going to go out on a limb and say that every person who has been on big companies online shopping site has seen this. At least anyone who has bought anything off of amazon.com. Collaborative filtering is a way for companies to try and get more sales. When you select a product to look at you often see a box on the side that shows you other products that people who bought what you are looking also bought. The idea is that you would possibly like what other people who bought what you bought, bought. In theory this sounds like a great idea, but there are flaws in it. What if a parent buys theirself things on amazon.com. But also does birthday and Christmas shopping too. If there were many parents buying similar things for themselves and for their kids, some of the suggestions might not be right for others.
Clark also describes a new type of search engine. A search engine that uses "transitiviy." This is creating links that are cause by people going from websites to websites. If a people go from website A, to B, to C, then there would be a link from site A to site C. There is a real problem with search engines. The web is such that ANYONE can put ANYTHING on it. Which means that there is a lot of JUNK out there. When searching for topics of many search engines it is hard to find exactly what you are looking for. You have to sift through a lor of junk. Making search engines better is a must.
It is interesting that Linux is not owned by anyone. I believe that Mac OS is based off of Linux. Mac to some extend also allows users to create amendments. The Widgets that Mac OSX has is taken from Linux. Microsoft also took it and used it on Vista because Mac OSX has made Widgets very popular. Microsoft, being original and creative (and a high quality system) has named theirs Gidgits.
Subscribe to:
Post Comments (Atom)
1 comment:
I agree that recommendation systems are interesting but have problems. I always think about the element of surprise when it comes to recommendation systems. If you base everything off of similarity, you may not get that many results that you like. For example, if I like Radiohead (which I do) then I might go through 10-15 recommended albums that are similar but I might get bored of that kind and want something new. Often I find music that that I would never expect to like--that I could never be recommended by Amazon.
The other problem is that there is still a lot of work to decide why you like something. Maybe I like a movie because of the directing. Or I like it because of the soundtrack or a specific actress. Recommendation systems can't really account for this type of phenomena.
Post a Comment