Similarity Forum

General Category => Wishlist => Topic started by: Spock on February 14, 2016, 00:54:07

Title: Improve The "Rating" Algorithm
Post by: Spock on February 14, 2016, 00:54:07
Hi,

I have a lot of duplicate songs (Apple Lossless codec) which come from different CD's and are mostly the same (they only vary by a few seconds) but they differ in audio quality. At present the "Rating" algorithm is not sensitive enough to tell the difference in quality between some songs - for example I have two songs which are the same in length and content, but the frequency and loudness vary quite considerably (to my ears). The rating assigned is "100%" to both, but one song is from an 80's CD and is too quiet and has a 'dull' sound because it lacks higher frequencies, while the other is from a more recent CD and has better loudness (without clipping) and improved treble, so better quality to my ears. But because both are assigned a rating of 100%, the Automark function can't distinguish between them and I have to mark them manually.

It would be nice if the Rating could be adjusted to be more sensitive to differences between loudness and frequency response? Maybe allowing the user to control the Rating sensitivity could be a useful addition, because at the moment I can't get any help from the Automark function and have to examine every duplicate song very carefully to avoid deleting the better quality files and it is taking me a long time.

Thanks for listening.
Title: Re: Improve The "Rating" Algorithm
Post by: Admin on February 29, 2016, 16:45:37
The new version of Similarity has possibility to implement own rating calculation algorithm with JavaScript (http://www.similarityapp.com/scripting), see analysis.js. But source values to rating can be used only from other analysis tab fileds.