<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
		>
<channel>
	<title>Comments on: More progress on soft tags</title>
	<atom:link href="http://mesquilla.com/2009/01/10/more-progress-on-soft-tags/feed/" rel="self" type="application/rss+xml" />
	<link>http://mesquilla.com/2009/01/10/more-progress-on-soft-tags/</link>
	<description>Messaging with Mozilla by rkent</description>
	<lastBuildDate>Sun, 18 Jul 2010 18:30:57 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.0</generator>
	<item>
		<title>By: rkent</title>
		<link>http://mesquilla.com/2009/01/10/more-progress-on-soft-tags/comment-page-1/#comment-14</link>
		<dc:creator>rkent</dc:creator>
		<pubDate>Thu, 05 Feb 2009 17:09:44 +0000</pubDate>
		<guid isPermaLink="false">http://mesquilla.com/?p=145#comment-14</guid>
		<description>I don&#039;t come at this with a personal &quot;long history&quot; in Bayesian classification, and the term &quot;feature&quot; is pretty overloaded in the computer world. Even in the example you gave, &quot;feature&quot; has one meaning in the Bayes world, and a different in the spam world. I was trying to select a more neutral term. But had I known that &quot;feature&quot; was a well-understood term in the Bayesian nomenclature, I probably would have left it.

Sorry about that ...</description>
		<content:encoded><![CDATA[<p>I don&#8217;t come at this with a personal &#8220;long history&#8221; in Bayesian classification, and the term &#8220;feature&#8221; is pretty overloaded in the computer world. Even in the example you gave, &#8220;feature&#8221; has one meaning in the Bayes world, and a different in the spam world. I was trying to select a more neutral term. But had I known that &#8220;feature&#8221; was a well-understood term in the Bayesian nomenclature, I probably would have left it.</p>
<p>Sorry about that &#8230;</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: era</title>
		<link>http://mesquilla.com/2009/01/10/more-progress-on-soft-tags/comment-page-1/#comment-13</link>
		<dc:creator>era</dc:creator>
		<pubDate>Thu, 05 Feb 2009 11:53:46 +0000</pubDate>
		<guid isPermaLink="false">http://mesquilla.com/?p=145#comment-13</guid>
		<description>(Found my way here from https://bugzilla.mozilla.org/show_bug.cgi?id=181866 -- thanks for the link.)

In https://bugzilla.mozilla.org/show_bug.cgi?id=181866#c57 you write:
&gt; I&#039;ve done a number of patches
&gt; that allow access to the bayesian filter
&gt; for features that are different than
&gt; junk, but I use the term &quot;traits&quot; rather
&gt; than &quot;features&quot;.

&quot;Traits&quot;?  The use of the term &quot;features&quot; in Bayesian classification and classification theory in general has a long history; moving away from that seems misdirected.  However, it seems that you are using &quot;trait&quot; as what is normally called &quot;category&quot;.  In spam classification the presence of a particular &quot;token&quot; is what is called a &quot;feature&quot;.</description>
		<content:encoded><![CDATA[<p>(Found my way here from <a href="https://bugzilla.mozilla.org/show_bug.cgi?id=181866" rel="nofollow">https://bugzilla.mozilla.org/show_bug.cgi?id=181866</a> &#8212; thanks for the link.)</p>
<p>In <a href="https://bugzilla.mozilla.org/show_bug.cgi?id=181866#c57" rel="nofollow">https://bugzilla.mozilla.org/show_bug.cgi?id=181866#c57</a> you write:<br />
&gt; I&#8217;ve done a number of patches<br />
&gt; that allow access to the bayesian filter<br />
&gt; for features that are different than<br />
&gt; junk, but I use the term &#8220;traits&#8221; rather<br />
&gt; than &#8220;features&#8221;.</p>
<p>&#8220;Traits&#8221;?  The use of the term &#8220;features&#8221; in Bayesian classification and classification theory in general has a long history; moving away from that seems misdirected.  However, it seems that you are using &#8220;trait&#8221; as what is normally called &#8220;category&#8221;.  In spam classification the presence of a particular &#8220;token&#8221; is what is called a &#8220;feature&#8221;.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
