Whose personality do you want today?
Fri Nov 21 17:47:08 GMT 2003 Permalink
l.m.orchard commented regarding the using Bayesian analysis on news. In fact, as soon as I saw it I remembered, I had read his piece already. It was probably his writing that triggered my initial interest in using a Bayesian classifier in K-Collector.
Re-reading that piece I got an interesting different angle since his approach was to blend a Bayesian classifier with his news aggregator to try and have it prioritize news he would find interesting and not to categorize it by topic. I think this is a much more scalable task, from a K-Collector perspective, than what Jon is experimenting with. I think the efforts of training a system-wide recognizer to differentiate between topics would be too much for most users of the product to bear.
Our product roadmap for K-Collector already includes allowing users to personalize the system. For example we think that people should be able to say which feeds they think are relevant on different topics. Notice that this is a much very granular relationship since it means that I can say "Matt Mower is a real expert on the topic sock puppets" but that this says nothing about how relevant I am on "dating." or any other topic. Indeed each user might rate the exact same sources differently over a wide range of topics.
What might be interesting is if people could "share" and "subscribe to" preference maps. As a new user of the system you might not really know who is relevant on any particular topic. But imagine you worked with David Weinberger, Phil Wolff, or Dan Gillmor. If you knew them and trusted their judgement you could pick one of their preference maps as a starting point and immediately gain a usseful insight into the data as it is structured by topic. You might even switch between personalities to get more perspective!
Thanks to l.m.'s piece I am now wondering also about whether a Bayesian classifier might be more use in helping users to establish their own preference maps about which content is most relevant to them.
Re-reading that piece I got an interesting different angle since his approach was to blend a Bayesian classifier with his news aggregator to try and have it prioritize news he would find interesting and not to categorize it by topic. I think this is a much more scalable task, from a K-Collector perspective, than what Jon is experimenting with. I think the efforts of training a system-wide recognizer to differentiate between topics would be too much for most users of the product to bear.
Our product roadmap for K-Collector already includes allowing users to personalize the system. For example we think that people should be able to say which feeds they think are relevant on different topics. Notice that this is a much very granular relationship since it means that I can say "Matt Mower is a real expert on the topic sock puppets" but that this says nothing about how relevant I am on "dating." or any other topic. Indeed each user might rate the exact same sources differently over a wide range of topics.
What might be interesting is if people could "share" and "subscribe to" preference maps. As a new user of the system you might not really know who is relevant on any particular topic. But imagine you worked with David Weinberger, Phil Wolff, or Dan Gillmor. If you knew them and trusted their judgement you could pick one of their preference maps as a starting point and immediately gain a usseful insight into the data as it is structured by topic. You might even switch between personalities to get more perspective!
Thanks to l.m.'s piece I am now wondering also about whether a Bayesian classifier might be more use in helping users to establish their own preference maps about which content is most relevant to them.
[