<?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: Music Search: Exploration or Lookup?</title>
	<atom:link href="http://irgupf.com/2009/10/21/music-search-exploration-or-lookup/feed/" rel="self" type="application/rss+xml" />
	<link>http://irgupf.com/2009/10/21/music-search-exploration-or-lookup/</link>
	<description>Information Retrieval Research, Issues, and Discussion</description>
	<lastBuildDate>Tue, 14 Jun 2011 14:26:42 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.1.3</generator>
	<item>
		<title>By: Information Retrieval Gupf &#187; Doing to Music What They Did to the Web</title>
		<link>http://irgupf.com/2009/10/21/music-search-exploration-or-lookup/comment-page-1/#comment-5276</link>
		<dc:creator>Information Retrieval Gupf &#187; Doing to Music What They Did to the Web</dc:creator>
		<pubDate>Thu, 29 Oct 2009 00:48:06 +0000</pubDate>
		<guid isPermaLink="false">http://irgupf.com/?p=967#comment-5276</guid>
		<description>[...] Music Search: Exploration or Lookup?  [...]</description>
		<content:encoded><![CDATA[<p>[...] Music Search: Exploration or Lookup?  [...]</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: jeremy</title>
		<link>http://irgupf.com/2009/10/21/music-search-exploration-or-lookup/comment-page-1/#comment-5199</link>
		<dc:creator>jeremy</dc:creator>
		<pubDate>Wed, 21 Oct 2009 23:52:40 +0000</pubDate>
		<guid isPermaLink="false">http://irgupf.com/?p=967#comment-5199</guid>
		<description>Jonathan,

I totally agree.  In fact, I should have also put this link, above:

http://irgupf.com/2009/04/07/music-explaura-exploration-and-discovery-in-action/

See also Paul&#039;s blog, for lots more discussion in this area.

Current: http://musicmachinery.com/
Archive: http://blogs.sun.com/plamere/</description>
		<content:encoded><![CDATA[<p>Jonathan,</p>
<p>I totally agree.  In fact, I should have also put this link, above:</p>
<p><a href="http://irgupf.com/2009/04/07/music-explaura-exploration-and-discovery-in-action/" rel="nofollow">http://irgupf.com/2009/04/07/music-explaura-exploration-and-discovery-in-action/</a></p>
<p>See also Paul&#8217;s blog, for lots more discussion in this area.</p>
<p>Current: <a href="http://musicmachinery.com/" rel="nofollow">http://musicmachinery.com/</a><br />
Archive: <a href="http://blogs.sun.com/plamere/" rel="nofollow">http://blogs.sun.com/plamere/</a></p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jonathan</title>
		<link>http://irgupf.com/2009/10/21/music-search-exploration-or-lookup/comment-page-1/#comment-5198</link>
		<dc:creator>Jonathan</dc:creator>
		<pubDate>Wed, 21 Oct 2009 22:49:49 +0000</pubDate>
		<guid isPermaLink="false">http://irgupf.com/?p=967#comment-5198</guid>
		<description>You&#039;re right that the most interesting music searches aren&#039;t known item searches.  If I want an mp3 of something I already know, then text search is fine.  Even if I don&#039;t know the song, but have a recording of it, acoustic fingerprinting (apparently) works well enough.

I often want new music, which is by definition an unknown/exploratory search problem.  Recommender systems like Pandora sometimes work, but I&#039;ve found (at least with Pandora) they suffer from lack of diversity or sometimes just put correlated musicians together that don&#039;t actually sound anything alike.  (Right now Pandora suggested the Silversun Pickups on my TV on the Radio station.  Both bands I enjoy immensely, but I wouldn&#039;t characterize them as similar.)  Trying to direct your search with a recommender is like playing blind man&#039;s bluff.  If I thumb up a song, I&#039;m saying I like the song, which I might, but I&#039;m also saying &quot;show me more of this,&quot; which isn&#039;t necessarily what I&#039;m looking for.  When I thumb one down, I&#039;m saying &quot;don&#039;t show me this,&quot; but what if I do like the song, it&#039;s just not what I&#039;m looking for right now?

It would be nicer if instead of just text search and recommender systems, we had a system that took text search as a starting point (say a song title, artist, or genre) and then let user explore not only the social/historical connections among the musicians but also navigate the style information (the &quot;music genome&quot; in Pandora-speak).  Not even necessarily in music theory language (I have no idea what &quot;vamping&quot; is, but Pandora says I like it.), but just &quot;lighter,&quot; &quot;darker,&quot; or perhaps expressing (dis)similarity with respect to other musicians.</description>
		<content:encoded><![CDATA[<p>You&#8217;re right that the most interesting music searches aren&#8217;t known item searches.  If I want an mp3 of something I already know, then text search is fine.  Even if I don&#8217;t know the song, but have a recording of it, acoustic fingerprinting (apparently) works well enough.</p>
<p>I often want new music, which is by definition an unknown/exploratory search problem.  Recommender systems like Pandora sometimes work, but I&#8217;ve found (at least with Pandora) they suffer from lack of diversity or sometimes just put correlated musicians together that don&#8217;t actually sound anything alike.  (Right now Pandora suggested the Silversun Pickups on my TV on the Radio station.  Both bands I enjoy immensely, but I wouldn&#8217;t characterize them as similar.)  Trying to direct your search with a recommender is like playing blind man&#8217;s bluff.  If I thumb up a song, I&#8217;m saying I like the song, which I might, but I&#8217;m also saying &#8220;show me more of this,&#8221; which isn&#8217;t necessarily what I&#8217;m looking for.  When I thumb one down, I&#8217;m saying &#8220;don&#8217;t show me this,&#8221; but what if I do like the song, it&#8217;s just not what I&#8217;m looking for right now?</p>
<p>It would be nicer if instead of just text search and recommender systems, we had a system that took text search as a starting point (say a song title, artist, or genre) and then let user explore not only the social/historical connections among the musicians but also navigate the style information (the &#8220;music genome&#8221; in Pandora-speak).  Not even necessarily in music theory language (I have no idea what &#8220;vamping&#8221; is, but Pandora says I like it.), but just &#8220;lighter,&#8221; &#8220;darker,&#8221; or perhaps expressing (dis)similarity with respect to other musicians.</p>
]]></content:encoded>
	</item>
</channel>
</rss>

