Comments on: Order, Part 1: CD Collections http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/ Technology. Code. Living relentlessly in the real world. Wed, 04 Aug 2010 04:02:24 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Gary Fleming http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-338 Gary Fleming Fri, 22 Apr 2005 12:52:53 +0000 <p>The first and last are certainly possible with Winamp 5 (the latter would have to be backed by some form of web service, I would imagine.) The second one I'm not sure I completely understand, but imagine it would be possible with a lot of work.</p><p>Also, no underhand knocking of <cite>13</cite>, Blur's best album to date.</p> The first and last are certainly possible with Winamp 5 (the latter would have to be backed by some form of web service, I would imagine.) The second one I’m not sure I completely understand, but imagine it would be possible with a lot of work.

Also, no underhand knocking of 13, Blur’s best album to date.

]]> By: Derek http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-339 Derek Thu, 21 Apr 2005 16:59:38 +0000 <p>Heh, you know that my preference is for hand-waving abstractions over specific operations. (Many of these would be predicated on a more-complete listener history, which records every individual listen, and—perhaps—track-skip) Let's see:</p><ul><li>Standard folksonomical zeitgeist display, parameterised by time period (e.g. "last 7 days" or "between 2100 and 0000").</li><li>Impressionistic playlist generation. Either graph-based, as suggested in my previous comment, or something more novel. My very vague idea for this would be a 2-d region (the dimensions being time and probability) on which it would be possible to "paint" tags using tools familiar to users of every paint package since MacPaint. Each tag that participates explicitly in the playlist would be represented by a colour (along with complete randomness, which would be transparent). The playlist could then be generated with the probability of a tag being chosen at any one time being equal to the proportion of pixels that are set to that tag's colour at that point on the time axis. I could describe this more intuitively, but it would require images.</li><li>Collaborative tagging. Services like <a href="http://www.audioscrobbler.com/">AudioScrobbler</a> appeal to me for the aggregation possibilities. Currently, it makes recommendations based on the similarity between your listening habits and those of others: the use of tags could aid this. Say, for example, that you were a fan of Blur in the mid-1990s. Then the listening habits of someone who listens to a lot of Blur that is tagged with <code>britpop</code> would carry more weight than those of someone who listens to <cite>13</cite> on repeat (tagged with <code>inaccessible</code>). The collaborative aspects of such a scheme interest me, with the possibility that they would save users from having to enter their own tags for bands or songs (though there would be some need for the ability to explicity delete a collaborative tag, if, for example, you didn't believe that The Darkness are <code>aces</code>).</li></ul><p>Just a few ideas, based on a couple of minutes' thought. I dare say that more imaginative uses could come to the fore if it were possible.</p> Heh, you know that my preference is for hand-waving abstractions over specific operations. (Many of these would be predicated on a more-complete listener history, which records every individual listen, and—perhaps—track-skip) Let’s see:

  • Standard folksonomical zeitgeist display, parameterised by time period (e.g. “last 7 days” or “between 2100 and 0000″).
  • Impressionistic playlist generation. Either graph-based, as suggested in my previous comment, or something more novel. My very vague idea for this would be a 2-d region (the dimensions being time and probability) on which it would be possible to “paint” tags using tools familiar to users of every paint package since MacPaint. Each tag that participates explicitly in the playlist would be represented by a colour (along with complete randomness, which would be transparent). The playlist could then be generated with the probability of a tag being chosen at any one time being equal to the proportion of pixels that are set to that tag’s colour at that point on the time axis. I could describe this more intuitively, but it would require images.
  • Collaborative tagging. Services like AudioScrobbler appeal to me for the aggregation possibilities. Currently, it makes recommendations based on the similarity between your listening habits and those of others: the use of tags could aid this. Say, for example, that you were a fan of Blur in the mid-1990s. Then the listening habits of someone who listens to a lot of Blur that is tagged with britpop would carry more weight than those of someone who listens to 13 on repeat (tagged with inaccessible). The collaborative aspects of such a scheme interest me, with the possibility that they would save users from having to enter their own tags for bands or songs (though there would be some need for the ability to explicity delete a collaborative tag, if, for example, you didn’t believe that The Darkness are aces).

Just a few ideas, based on a couple of minutes’ thought. I dare say that more imaginative uses could come to the fore if it were possible.

]]> By: Gary Fleming http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-340 Gary Fleming Wed, 20 Apr 2005 09:57:24 +0000 <p>What specific non-search operations do you mean?</p><p>Winamp at least provides a healthy plugin architecture from which your wishes could be fulfilled.</p> What specific non-search operations do you mean?

Winamp at least provides a healthy plugin architecture from which your wishes could be fulfilled.

]]> By: Derek http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-341 Derek Tue, 19 Apr 2005 21:20:47 +0000 <p>And I could also do that in iTunes, but it seems that neither gives me the wherewithal to perform non-search operations on the tags. So my dreams of an MP3 zeitgeist, for example, go unfulfilled. Or a playlist, generated based on tags, that takes me on a path from <i>twee</i> to, umm, <i>hardcore</i>. Or a million different graph-drawing possibilities.</p> And I could also do that in iTunes, but it seems that neither gives me the wherewithal to perform non-search operations on the tags. So my dreams of an MP3 zeitgeist, for example, go unfulfilled. Or a playlist, generated based on tags, that takes me on a path from twee to, umm, hardcore. Or a million different graph-drawing possibilities.

]]> By: Gary Fleming http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-342 Gary Fleming Tue, 19 Apr 2005 14:40:01 +0000 <p>Tagging mechanism? Why, you could add tags to the comment section of the MP3 and then let the WinAmp 5 media library and its querying functions deal with the rest.</p> Tagging mechanism? Why, you could add tags to the comment section of the MP3 and then let the WinAmp 5 media library and its querying functions deal with the rest.

]]> By: Derek http://solitude.vkps.co.uk/Archives/2005/04/16/OrderPart1CDCollections/comment-page-1/#comment-343 Derek Tue, 19 Apr 2005 13:19:23 +0000 <p>When I kept my CDs in some semblance of order, it was in the same manner as your original scheme. Now, I just pile them into boxes in any old order: shameful, I know, but when every track is loaded on my computer, I don't need easy access to the original media. (What I want but don't yet have is an elegant folksonomical tagging mechanism for the media library... any suggestions?)</p><p>I like the 2-D arrangement idea though. The algorithmicist (read: geek) in me would be intrigued to know if there could be some automatic way of determining an optimum layout, or if this is a completely intractable problem. (My thinking involves conditional probability: what is the probability that you'd listen to another artist, given that you had just listened to a particular artist. Where the probability is high, the two artists should maybe be close together. Of course, this assumes that you only listen to music of the same genre at one sitting. Still, a more proficient AI practitioner than me could probably come up with something.)</p> When I kept my CDs in some semblance of order, it was in the same manner as your original scheme. Now, I just pile them into boxes in any old order: shameful, I know, but when every track is loaded on my computer, I don’t need easy access to the original media. (What I want but don’t yet have is an elegant folksonomical tagging mechanism for the media library… any suggestions?)

I like the 2-D arrangement idea though. The algorithmicist (read: geek) in me would be intrigued to know if there could be some automatic way of determining an optimum layout, or if this is a completely intractable problem. (My thinking involves conditional probability: what is the probability that you’d listen to another artist, given that you had just listened to a particular artist. Where the probability is high, the two artists should maybe be close together. Of course, this assumes that you only listen to music of the same genre at one sitting. Still, a more proficient AI practitioner than me could probably come up with something.)

]]>