What is collaborative filtering?
What follows is a reflection on "collaborative filtering". The content of this reflection is essentially a paraphrase of the pages 143 to 148 of Chapter 6 - Slugs, Ants and Amazon.com - by Andy Clark (2003): Natural-Born Cyborgs , Oxford University Press.
Browsing the Web leave tracks.
The electronic traces we leave when we access data, when we use a search engine, when we make a purchase online, when we communicate, can be tracked, analyzed, grouped together with traces of other Web surfers, and ultimately, exploited to generate new knowledge.
A special use of this new knowledge generated by the interaction of the traces left on the net is what make sites like http://www.amazon.com , one of the biggest and most popular commercial sites. Amazon.com uses tracks electronic leave users who visit your site for information and customize the report maintains that with its customers.
Amazon.com is the prime example of how it can be harnessed "knowledge" that we allow the network to develop a quality site that can customize the information that it transmits.
me give an example. We intend to buy a CD of Peter Tosh through Amazon. While we buy that cd we are informed: 'People who bought this album also bought ...'. We are informed that a series of CDs has been purchased by buyers of the same cd of Peter Tosh. This process is commercially very effective. In fact often happens that a person will find interesting cd did not know, and that match your tastes, following the trails of other purchasers of the same cd that is likely to have the same interests.
The IT architecture of Amazon, using the traces left by its customers, is based on dynamic knowledge structures, capable of self-organization (self organizing knowledge structures).
More specifically, the technique used to suggest a CD to a user of the Amazon depends on a system of collaborative filtering ( collaborative filtering). Collaborative filtering systems to propose a custom tips through the computation similarities between our preferences and those of other people.
As mentioned above, each share of a buyer leaves traces in the Amazon. After a sufficient number of shares similar patterns emerge in the web of choices shared by multiple buyers. Suppose that I, the buyer of the CD of Peter Tosh, also includes two other CD. One of the CD is a gift for a friend with musical tastes are very different from mine, the second is another CD I buy for myself. Likely choices from the second cd will be not much different from the choices of other fans of Peter Tosh who buy CDs in the Amazon. In this case, the electronic traces, organized according to the technique of collaborative filtering, will lead other buyers of the CD by Peter Tosh to products that meet their taste in music.
The simple tactic to ensure that the buyer's online activities leave electronic traces that can be followed by other buyers reveals a powerful type of cognitive structure. One such procedure is powerful because it allows the action patterns of buyers' speak for itself "and will signify in the space of Web paths immediate and municipalities. These collective paths have the advantage of delineating categories are not ordinary. These categories emerge automatically deviate from the categories we use to routinely rank our choices. The category, in this case, the activity is created directly Buyers Online: It is a type of category planning, emerges from collective choices and thus so flexible as the choices of buyers.
Another advantage of a system based on collaborative filtering. Imagine a system that not only keeps track of what different people have chosen, but also the temporal sequence of choices. Such a system could develop common pattern of evolution of the choices and thus offer useful advice on what to try to choose in the future.
* Learn About Collaborative Filtering:
http://www.sims.berkeley.edu/resources/collab/
A resource website School of Information Management & Systems at UC Berkeley, CA.
http://jamesthornton.com/cf/
Links to papers on Collaborative Filtering, Internet consultant from the site of James Thornton.
http://pespmc1.vub.ac.be/COLLFILT.html
Resource from the website of the International Principia Cybernetica Project (PCP).
http://en.wikipedia.org/wiki/Collaborative_filtering
Wikipedia on Collaborative Filtering.
http://www.html.it/architettura_informazione/infoarch_29.htm
Brief article that explains what the Italian collaborative filtering. From HTML.It.
* Some commercial sites that use the system of collaborative filtering:
http://www.amazon.com/
http://www.half.ebay.com/
http://www.barnesandnoble.com/
http://www.musicmatch.com/
http://www.netflix.com/
For other links to sites that use collaborative filtering technology to link the wikipedia reference above.
0 comments:
Post a Comment