[Hpr] HPR wishlist topic request

dg dg at deepgeek.us
Mon Jun 23 17:24:12 PDT 2014


Thanks for reply, Charles,

A long time ago, I was on a lot of mailing lists and used "popfile"
which, instead of doing bayes on spam/nonspam, did as many designations
as you wanted.  I did spam/boring/interesting.  It got pretty good at
figuring out what kind of discussion I really liked to see.  The idea
of doing this for RSS feed entries sounds really great to me!  I have
many podcast feeds in my reader, and I read new entries to pick ten
that I want to download at a time, training a program to key in on my
tastes for me sounds like a time saver.  BTW, did you ever read the
William Gibson novel Idoru?  The AI called "the music man" knows her
owners tastes.  Sounds simular.  

Actually, when I wrote this, I thinking more of a government doing some
great firewall thing to email discussion, or some government finding
dissident webpages with this technology.  Yet we can dream up good
uses.  Goes to show that the tool is not moral/immoral, rather the user
of the tool is.

Thanks!
---
DeepGeek

On Sun, 22 Jun 2014 16:35:14 -0700
Charles Thayer <catintp at yahoo.com> wrote:

> Wow! That sounds like a cool topic.
> 
> My first instinct would be to try to use naive Bayesian classifiers
> to make a news headlines feed that works better and better over time,
> based on user feedback (Like/Dislike, or maybe degree of Like).  The
> only side effect of doing this would be a self-imposed "filter
> bubble" closed system that could develop over time due to my own
> preferences.  To cure this, I would want to introduce some randomness
> here and there to introduce new types of entries.
> 
> Actually, that sounds like a news feed version of Pandora.
> 
> 
> Great idea.
> 
> Charles in NJ
> 
> 
> 
> ________________________________
>  From: dg <dg at deepgeek.us>
> To: hpr at hackerpublicradio.org 
> Sent: Sunday, June 22, 2014 1:50 PM
> Subject: [Hpr] HPR wishlist topic request
>  
> 
> After watching...
> 
> https://www.youtube.com/watch?v=JKB5CojW4AA
> 
> wanted to request the topic "alternative uses for Bayesian email
> classifiers"....
> 
> ---
> DeepGeek
> 
> _______________________________________________
> Hpr mailing list
> Hpr at hackerpublicradio.org
> http://hackerpublicradio.org/mailman/listinfo/hpr_hackerpublicradio.org





More information about the Hpr mailing list