tag:help.beatunes.com,2009-07-24:/discussions/problems/49876-mood-analysis-accuracybeaTunes: Discussion 2019-03-03T16:53:25Ztag:help.beatunes.com,2009-07-24:Comment/470219472019-03-01T16:20:28Z2019-03-01T16:20:28ZMood analysis accuracy<div><p>(ok, so the original file went through anyway, it seems.)</p></div>victortag:help.beatunes.com,2009-07-24:Comment/470219472019-03-02T08:38:52Z2019-03-02T08:38:52ZMood analysis accuracy<div><p>Hey Victor,</p>
<blockquote>
<p>Are you doing your own mood analysis or do you only rely on last.fm?</p>
</blockquote>
<p>beaTunes 5.x does the following:</p>
<ol>
<li>The beaTunes database is asked. If a mood is found, it is used and processing stops.<br></li>
<li>Last.fm is asked. If the track is found, tags are scanned for emotionally significant words and valence/arousal values are inferred. Those are then used and processing is stopped.<br></li>
<li>As a last resort, AcousticBrainz is asked and its mood values are used.</li>
</ol>
<p>The beaTunes database consists of values submitted by users. Those values in turn may have been inferred from Last.fm tags, from AcousticBrainz, or have been manually set. What's currently returned by the beaTunes database server is the median of both the stored valences and arousals, regardless of how they were produced.</p>
<p>Cheers,</p>
<p>-hendrik</p></div>hendriktag:help.beatunes.com,2009-07-24:Comment/470219472019-03-03T16:43:30Z2019-03-03T16:43:30ZMood analysis accuracy<div><p>No word on why that one track is mis-characterized?</p>
<p>Anyway, I was hoping you were doing some fancy DSP’ing that could be tweaked.</p>
<p>V.</p></div>victortag:help.beatunes.com,2009-07-24:Comment/470219472019-03-03T16:47:36Z2019-03-03T16:47:36ZMood analysis accuracy<div><blockquote>
<p>No word on why that one track is mis-characterized?</p>
</blockquote>
<p>No, sorry, but it would not change anything.</p>
<blockquote>
<p>Anyway, I was hoping you were doing some fancy DSP’ing that could be tweaked.</p>
</blockquote>
<p>The alternative way to using user content (Last.FM tags) is to do machine learning on user content... This would probably improve results for items that <em>cannot</em> be found on Last.FM.</p></div>hendriktag:help.beatunes.com,2009-07-24:Comment/470219472019-03-03T16:50:50Z2019-03-03T16:50:50ZMood analysis accuracy<div><p>Or plain fancy DSP’ing. You’re already doing some of that to determine the volume. It must be possible to throw more Fourier analysis against it to determine mood and such. ML can play a part in that.</p>
<p>Victor.</p></div>victortag:help.beatunes.com,2009-07-24:Comment/470219472019-03-03T16:53:24Z2019-03-03T16:53:24ZMood analysis accuracy<div><p>The current thinking in the scientific community is supervised learning. As NN input you'd typically use Mel spectrograms. No fancy DSPing necessary. That's the beauty of machine learning.</p>
<p>I'm totally interested in doing this... but it won't happen for v5. Hopefully v6. We'll see. There is only so much I can do (time is the issue here).</p></div>hendrik