AutoRateAccurate - Release 1.61 (10/11/2006) [MM2]

Big_Berny · Post by **Big_Berny** » Sun Apr 01, 2007 8:18 am

Hey lemo,
cool that you like it.

Yes, I want to develop it further but at the moment I don't have as much time and I also wasn't to motivated. But there will surely be a new version sometimes. So yes, I'm interested in your ideas!

Big_Berny

lemo · Post by **lemo** » Sun Apr 01, 2007 9:59 am

OK here goes...

Currently, self-rated songs are recognised because their rating is a multiple of 10. Auto-rated tracks end have values ending in 5. You can look at this slightly differently by saying that self-rated songs have even-numbered ratings while auto-rated songs are odd.

If use odd/even to recognise the difference, we can alter the ratings of self-rated songs while keeping a memory of the fact that they were self-rated by making sure that changes to their ratings always result in another even number.

We can then have two formulae, one that calculates an auto-rating for songs that have never been self-rated (no change there) and a second one that calculates deltas that are a multiple of two for self-rated songs.

Self-ratings should be quite "sticky" so these deltas should be quite small (2 or 4 maybe) but over time would result in the rating drifting upwards or downwards depending on skips and so on.

This is the basic idea and there are still some issues that I can think of:

- What formula makes sense for calculating the delta?
- How do we stop the delta from being applied each time AutoRateAcc runs when nothing has changed for the song since last time? This might be a big problem, in fact.

Still, I would be interested in your thoughts...

Big_Berny · Post by **Big_Berny** » Sun Apr 01, 2007 1:23 pm

Hey lemo,
that sounds quite interesting. But first an info: Autorated songs don't have values ending in 5 in all cases! I always use the whole range of ratings.

Example:
To get 4 stars a rating has to be in the range of 76-85. Autorated songs now normally get 85 like you said. But with every skip the rating will decrease by 1. So a song with 84 has 4 stars and 1 skip, a song with 81 has 4 stars but 4 skips. So I use the ratnig field for two informations: The rating and the SkipCounter. Because of this I don't know if it's a good idea to use odd/even numbers. But I'm still waiting for MM 3 because then there is maybe a better possibility to use additional fields. Then I don't have to store the SkipCounter in the rating field anymore what gives me more possibilities like adding a "delta-formula" like you said.

lemo · Post by **lemo** » Sun Apr 01, 2007 4:18 pm

Ah, I missed that trick but it makes sense. Pity, alright. Looks like we'll have to hope MM3 helps somehow.

tooscaredtotry · Post by **tooscaredtotry** » Tue Apr 03, 2007 5:46 pm

Big_Berny wrote:
tooscaredtotry wrote: Ahhh, so that's why it gives me multiple entries in the built in library stats. Does anyone have a fix for this without installing another script..
What's exactly your problem?

when i go to edit, create report or whatever it's called .. the listing will show three entries for say 4 stars, each with a different total.. which makes total sense because the ratings numerical value is actually different.. i guess it just needs a patch to the advanced report script that accounts for this.

i'll take a screenshot later to show.

Big_Berny · Post by **Big_Berny** » Tue Apr 03, 2007 6:56 pm

tooscaredtotry wrote:when i go to edit, create report or whatever it's called .. the listing will show three entries for say 4 stars, each with a different total.. which makes total sense because the ratings numerical value is actually different.. i guess it just needs a patch to the advanced report script that accounts for this.

i'll take a screenshot later to show.

Ah, I understand now. You should inform Jiri or Rusty about that so they can fix it.

Big_Berny

tooscaredtotry · Post by **tooscaredtotry** » Tue Apr 03, 2007 10:25 pm

Here's a screenshot too

Spazz · Post by **Spazz** » Thu Apr 05, 2007 3:57 pm

Are you planning to update the script so it will work with MM3? Or waiting till it's out of alpha?

The only things keeping me from using MM3 right now is the lack of a DB import and the autorating script. I soooo wanna use it because you can nest smart playlists now but I can't

Big_Berny · Post by **Big_Berny** » Thu Apr 05, 2007 6:18 pm

Yes, I wait for MediaMonkey 3.0 Beta and then I'll post an update with some new features: planned is a 'base ranking' for songs which you never listened. So unlistened song all get 3 stars for example. And another feature I want to implement is a changeble formula or at least the possibility to change the weight of SkipPerDay.

By the way when an alpha is available you can use it without AutoRateAccurate first. Because the script can recalculate the rating anyway if you install it afterwards.

Big_Berny

Spazz · Post by **Spazz** » Fri Apr 06, 2007 11:18 am

That is true, but I also need my old library imported in, unless there is a way I don't know about.

Big_Berny · Post by **Big_Berny** » Fri Apr 06, 2007 12:57 pm

Yes, your're right. The current alpha don't import the library. We have to wait for a Beta verion I think...

Big_Berny · Post by **Big_Berny** » Sun May 13, 2007 6:59 pm

Hey,
I finally have some new ideas I'd like to share!

AutoRateAccurate 2.0 (Codename: Revolution)
Since some months I have a statistic module in my study. And this gave me the basic idea for version 2 of my script. I'd like to introduce a new feature called "formula generator". As the name already says this function will have to possibility to automatically create a formula - based on your own listening habbit.
How it works: Probably the script will show some songs of your library and then you have to rate them. The script then tries to 'understand' how you set the ratings, it will calculate the importance/weight of the different variables like SkipCounter, PlayedCounter, DaysSincePlayed. So you'll get a 'perfect' formula just for you, based on your reference ratings.
This part will probably need a lot of code but I assure that the script won't get slower.

What do you think of that idea?

BTW since some months I didn't have a lot time for working on that script. And I can't say when 2.0 will be final but this idea gives me some new motivation. And if someone want's to help developing this script I'd be very grateful!

Regards
Big_Berny

PS: Please forgive me the "Codename: Revolution" but I study a kind of public relations and marketing so I like these therms.

Spazz · Post by **Spazz** » Sun May 13, 2007 8:32 pm

I'm actually debating learning some scripting this summer before Uni starts again, good practice for my next CSCI classes. I got Computer Science 1 and 2 under my belt already, should be enough.

Only thing is before massive code changes, a workable version for MM3 would be nice, even if the DB converter isn't up yet.

lemo · Post by **lemo** » Mon May 14, 2007 3:39 am

I think it's an interesting thing to try but I would not be very confident of the accuracy of what you're going to get. The amount of data you have for each track is going to be very limited for anyone with a large database. So correlations will inevitably be weak. If I were you, I would look for a few volunteers and make a test script which doesn't alter the real ratings in the DB and get some feedback on how it works to see whether it's worthwhile taking it forward.

While I'm at it, do you think the MM3 database will allow anything to be done to allow self-rated songs to be modified using a different ("delta") formula?

Big_Berny · Post by **Big_Berny** » Mon May 14, 2007 11:40 am

@Spazz: I don't know if I'll release a 1.x version for MM3 because it also needs changes which make it incompatble to MM2.

@lemo: I don't think that correlation will be as weak. Because my script uses the point-ranking of songs to calculate the rating. And surely the formula won't be completely dynamic/flexible but a lot. Maybe I also implement for example 10 formulas and the functions looks which is the best for you.