<?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"
	>
<channel>
	<title>Comments on: The Two Big Problems For Powerset</title>
	<atom:link href="http://www.psynixis.com/blog/2008/05/12/the-two-big-problems-for-powerset/feed/" rel="self" type="application/rss+xml" />
	<link>http://www.psynixis.com/blog/2008/05/12/the-two-big-problems-for-powerset/</link>
	<description>Simon Brocklehurst's Technology Blog</description>
	<pubDate>Wed, 19 Nov 2008 22:53:39 +0000</pubDate>
	<generator>http://wordpress.org/?v=2.6.3</generator>
		<item>
		<title>By: simon</title>
		<link>http://www.psynixis.com/blog/2008/05/12/the-two-big-problems-for-powerset/#comment-368921</link>
		<dc:creator>simon</dc:creator>
		<pubDate>Mon, 12 May 2008 21:09:16 +0000</pubDate>
		<guid isPermaLink="false">http://www.psynixis.com/blog/?p=435#comment-368921</guid>
		<description>Hi Mark,  Thank you for commenting.

So, I agree with you: keyword-based searches that lead to your summaries of the Wikipedia and Freebase content are more promising; the results are certainly relevant to the query with keyword searches.

This is where it gets tricky though. One of the questions I suspect people will be asking is: how much value are you adding to the data?

For example, presenting the results of a query to Freebase possibly doesn't add too much value (i.e. your tabbed summaries that often appear at the top of a results page); and presenting the list of "hits" of the keyword from an index search of Wikipedia in the last section of the results page, possibly doesn't either (that is, the list of links).    That's because you can get broadly equivalent results from Freebase and Google, respectively.

The middle section "Factz we found", which appears in some results pages, is, I think, genuinely unique to Powerset. That is, the  "subject verb object" triplets.  People's mileage is going to vary with this part of your results page.  Without doubt, there are some interesting nuggets of information sometimes there; but there's quite a bit of noise too.

Having said all that, it's easily possible I'm being overly critical (I spend quite a lot of time thinking about adding genuine value to data in vertical search engines, because of some of the work we do, and that makes me look at these kind of things more closely than your typical user might).   Your results pages present all the information in an easy-to-digest way, and there is value in that.  And there is value in a type of meta search that gives equivalent results to searches of Freebase and Google combined.  At this point, it's just not obvious to me how *much* (time will tell!).  I wish you every success though!</description>
		<content:encoded><![CDATA[<p>Hi Mark,  Thank you for commenting.</p>
<p>So, I agree with you: keyword-based searches that lead to your summaries of the Wikipedia and Freebase content are more promising; the results are certainly relevant to the query with keyword searches.</p>
<p>This is where it gets tricky though. One of the questions I suspect people will be asking is: how much value are you adding to the data?</p>
<p>For example, presenting the results of a query to Freebase possibly doesn&#8217;t add too much value (i.e. your tabbed summaries that often appear at the top of a results page); and presenting the list of &#8220;hits&#8221; of the keyword from an index search of Wikipedia in the last section of the results page, possibly doesn&#8217;t either (that is, the list of links).    That&#8217;s because you can get broadly equivalent results from Freebase and Google, respectively.</p>
<p>The middle section &#8220;Factz we found&#8221;, which appears in some results pages, is, I think, genuinely unique to Powerset. That is, the  &#8220;subject verb object&#8221; triplets.  People&#8217;s mileage is going to vary with this part of your results page.  Without doubt, there are some interesting nuggets of information sometimes there; but there&#8217;s quite a bit of noise too.</p>
<p>Having said all that, it&#8217;s easily possible I&#8217;m being overly critical (I spend quite a lot of time thinking about adding genuine value to data in vertical search engines, because of some of the work we do, and that makes me look at these kind of things more closely than your typical user might).   Your results pages present all the information in an easy-to-digest way, and there is value in that.  And there is value in a type of meta search that gives equivalent results to searches of Freebase and Google combined.  At this point, it&#8217;s just not obvious to me how *much* (time will tell!).  I wish you every success though!</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Mark Johnson</title>
		<link>http://www.psynixis.com/blog/2008/05/12/the-two-big-problems-for-powerset/#comment-368614</link>
		<dc:creator>Mark Johnson</dc:creator>
		<pubDate>Mon, 12 May 2008 17:46:56 +0000</pubDate>
		<guid isPermaLink="false">http://www.psynixis.com/blog/?p=435#comment-368614</guid>
		<description>I agree entirely that promoting natural language searches is not the way to succeed.  Powerset tries to add value to keyword searches as well as NL queries (try "elton john" and look at the Factz, some cool stuff in there).  Note also that we bring our semantic technology into our enhanced Wikipedia articles with tools that help users to read the content.  Play around with us for awhile and then see what you think.

-mark, powerset product manager</description>
		<content:encoded><![CDATA[<p>I agree entirely that promoting natural language searches is not the way to succeed.  Powerset tries to add value to keyword searches as well as NL queries (try &#8220;elton john&#8221; and look at the Factz, some cool stuff in there).  Note also that we bring our semantic technology into our enhanced Wikipedia articles with tools that help users to read the content.  Play around with us for awhile and then see what you think.</p>
<p>-mark, powerset product manager</p>
]]></content:encoded>
	</item>
</channel>
</rss>
