<?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: Social media at its worst &#8211; post-mortem cyber-bullying</title>
	<atom:link href="http://www.nickarnett.net/2008/12/15/social-media-at-its-worst-post-mortem-cyber-bullying/feed/" rel="self" type="application/rss+xml" />
	<link>http://www.nickarnett.net/2008/12/15/social-media-at-its-worst-post-mortem-cyber-bullying/</link>
	<description>Social media analytics for decision-making</description>
	<lastBuildDate>Fri, 16 Jul 2010 16:45:53 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.0</generator>
	<item>
		<title>By: Walter Underwood</title>
		<link>http://www.nickarnett.net/2008/12/15/social-media-at-its-worst-post-mortem-cyber-bullying/comment-page-1/#comment-21</link>
		<dc:creator>Walter Underwood</dc:creator>
		<pubDate>Mon, 15 Dec 2008 19:12:05 +0000</pubDate>
		<guid isPermaLink="false">http://www.nickarnett.net/?p=78#comment-21</guid>
		<description>A rapid burst is an indicator by itself.

When a user gets five messages per day from the same three people suddenly gets 30 messages in 10 minutes from 30 people with half of them negative, that is an event worth flagging.

Traffic analysis has been pretty well-understood by the military intelligence folk for a few decades, including rapidly flagging a new pattern of messages. Banks use similar algorithms for fraud detection, so there are good unclassified methods.

The main problem is the cost of manual moderation for false positives. My wild guess is that this would take a few hours of data to recognize a burst, but it might be much less, maybe 30 minutes.</description>
		<content:encoded><![CDATA[<p>A rapid burst is an indicator by itself.</p>
<p>When a user gets five messages per day from the same three people suddenly gets 30 messages in 10 minutes from 30 people with half of them negative, that is an event worth flagging.</p>
<p>Traffic analysis has been pretty well-understood by the military intelligence folk for a few decades, including rapidly flagging a new pattern of messages. Banks use similar algorithms for fraud detection, so there are good unclassified methods.</p>
<p>The main problem is the cost of manual moderation for false positives. My wild guess is that this would take a few hours of data to recognize a burst, but it might be much less, maybe 30 minutes.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Nick Arnett</title>
		<link>http://www.nickarnett.net/2008/12/15/social-media-at-its-worst-post-mortem-cyber-bullying/comment-page-1/#comment-20</link>
		<dc:creator>Nick Arnett</dc:creator>
		<pubDate>Mon, 15 Dec 2008 18:13:26 +0000</pubDate>
		<guid isPermaLink="false">http://www.nickarnett.net/?p=78#comment-20</guid>
		<description>Good thoughts... The big challenge is figuring out how to make it happen fast.</description>
		<content:encoded><![CDATA[<p>Good thoughts&#8230; The big challenge is figuring out how to make it happen fast.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Walter Underwood</title>
		<link>http://www.nickarnett.net/2008/12/15/social-media-at-its-worst-post-mortem-cyber-bullying/comment-page-1/#comment-19</link>
		<dc:creator>Walter Underwood</dc:creator>
		<pubDate>Mon, 15 Dec 2008 17:53:54 +0000</pubDate>
		<guid isPermaLink="false">http://www.nickarnett.net/?p=78#comment-19</guid>
		<description>How about a comment hold when there is a burst of negative comments? The hold could be lifted by the owner or a moderator.

Sentiment extraction is getting better and probably good enough for this application. I&#039;m sure there are some other features to train on, maybe negative comments from first-time commenters.</description>
		<content:encoded><![CDATA[<p>How about a comment hold when there is a burst of negative comments? The hold could be lifted by the owner or a moderator.</p>
<p>Sentiment extraction is getting better and probably good enough for this application. I&#8217;m sure there are some other features to train on, maybe negative comments from first-time commenters.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
