Every key is limited to 1000 results on catalog read. If you need to read all the results, I would recommend using the start on catalog/read and use the "total" in the response to iterate over the entire taste profile.
↧
Check taste profile status
↧
Limited results for song radio?
My intent is eventually to use the taste profile API to provide the feedback in not-realtime. (I have no interest in writing a GUI at this point.) I realize there is a limitation with dynamic stations disappearing after a day of idleness, so it seems that the taste profile API (one per station) should work around that. Is that correct?
The lifecycle would be create dynamic station with seed, produce a playlist and a taste profile, save the profile id (and probably a list of song name, artist, track id) somewhere, go back and update the taste profile after some interval of listening, then create a new dynamic station seeded with only the taste profile, or possibly the original seed(s) plus the taste profile, and generate another big playlist, repeat.
Will this work? If so, should the new dynamic station be seeded with the taste profile only, or the original seeds plus the taste profile?
Thanks for the prompt responses by the way.
↧
↧
Limited results for song radio?
Taste Profiles are, pretty much by definition, how "personalization" and long-term memory of a user's preferences work in our world.
If I was trying to build this app, I'd probably use static playlists to get 100 song chunks, record songs that were listened to in a taste profile, and then get another 100 song chunks.
As a bit of a side note, you don't need to use spotify-WW for those ids you're feeding us; you can just straight-up use the spotify URI.
↧
Limited results for song radio?
Thanks for the response.
Regarding spotify URIs, do you mean this: track_id=spotify-WW:track:2EB1aFLWC4uIfld5Z3Yj1z vs this: track_id=2EB1aFLWC4uIfld5Z3Yj1z ?
If so, the latter didn't work when I tried it.
Getting static playlists in 100-song chunks with a taste profile -- will there be duplicates from one 100-song chunk to others? In the 2000 song playlist I generated, there were ZERO duplicates.
↧
Limited results for song radio?
No, track_id=spotify:track:2EB1aFLWC4uIfld5Z3Yj1z
If you record tracks you have listened to in the taste profile, use type=catalog-radio, and pass the adventurousness=1 parameter, we shouldn't repeat tracks.
↧
↧
Limited results for song radio?
Awesome. I love adventurousness.
Thanks for all the responses.
↧
Taste profile -- rosetta id or ENID, song vs. track
Hi. I have some questions about the taste profile.
The api reference at http://developer.echonest.com/docs/v4/tasteprofile.html#update says that track IDs can be rosetta ID or ENID. Is this a rosetta ID? spotify:track:2EB1aFLWC4uIfld5Z3Yj1z
I noticed that when I create a playlist, there is a one to many relationship between song and track. The results list a song, and associated with the song are many tracks (for example, recordings on different albums).
Is it more effective to update taste profiles using song_id rather than track_id, so as to have the information apply to all instances of the song (i.e., same song, different tracks in different albums) rather than just a single track (a single track in a single album), or is echo nest smart enough to take care of that on its own?
Thanks in advance.
↧
Taste profile -- rosetta id or ENID, song vs. track
"spotify:track:2EB1aFLWC4uIfld5Z3Yj1z" is, indeed, a Rosetta ID.
We'll automatically associate tracks in a taste profile with the appropriate song.
↧
Taste profile -- rosetta id or ENID, song vs. track
Thanks.
So banning a track has the effect of banning a song?
↧
↧
Taste profile -- rosetta id or ENID, song vs. track
(subscribe to email updates)
↧
Taste profile -- rosetta id or ENID, song vs. track
That's correct!
↧
Is there a way to get a list of EN song dataset
Hi Guys,
as per my thread title, is there a way to get a list of EN song dataset?
Cheers
↧
Is there a way to get a list of EN song dataset
Not sure what you mean. Could you be more specific?
-- Paul
↧
↧
Is there a way to get a list of EN song dataset
Hi Paul,
Thanks for the reply.
i am thinking if there is a way to get the song summary(valence & energy) of all songs available on EN.
↧
Is there a way to get a list of EN song dataset
Sorry no, we don't make full dumps of our data available. -- Paul
↧
TasteProfile problem updating (adding) songs
Hi
I'm trying to upload a music library to a taste profile. I upload songs in batches of 500. I get a response that says success that also includes an "update_info" array.
{ response = { "items_updated" = 500; "percent_complete" = 100; status = { code = 0; message = Success; version = "4.2"; }; "ticket_status" = complete; "total_items" = 500; "update_info" = ( { info = "couldn't resolve item"; "item_id" = 6152345113083139857; }, { info = "couldn't resolve item"; "item_id" = 6152345113083139858; }, { info = "couldn't resolve item"; "item_id" = 6152345113083139885; }, { info = "couldn't resolve item"; "item_id" = 6152345113083139887; }, . . . etc.
I usually get about 85 errors in 500 items. That's probably about right given my music collection, but when I read the taste profile, there are no items. Is this a bug? Shouldn't there be ~415 items? Or am I supposed to get the item_id of each update_info and then upload my entire library again without those items?
↧
Waveform API
It would be cool if the echonest API would also return a waveform (in JSON form preferably to allow custom styling) for a track. I'm not sure why this isn't a feature yet considering the synchstring feature but I can see this being very useful for everyone that wants to display the full waveform without having to have the whole audio data on disk beforehand (for example in a streaming environment).
There is already an opensource library that produces such json files from .wav files: https://github.com/liufengyun/waveformjson
It would be great if you'd consider adding this to your lineup of already awesome features.
↧
↧
Limited results for song radio?
If I was trying to build this app, I'd probably use static playlists to get 100 song chunks, record songs that were listened to in a taste profile, and then get another 100 song chunks. <<
I played with this a bit today.
If I have a 100-song playlist and I load all 100 songs into a taste profile, then generate a playlist using catalog-radio and specifying that taste profile, the results are unacceptable. The song selection is all over the place.
Generate a playlist using song-radio, and specify a catalog and a track. The catalog is a taste profile generated from a 100-song playlist that was created using song-radio with the same track specified for the new playlist. The results are more acceptable (i.e., the songs are similar to the seed song), but every song logged in the taste profile is repeated in the new playlist. Even with adventurousness=1 and variety=1, the tracks in the taste profile are repeated. This is not what you said you expected.
↧
TasteProfile problem updating (adding) songs
Actually, it looks like even if I do remove the songs that can't be resolved, I still get 0 items with tasteprofile/read. Same even if I upload a single song.
↧
Limited results for song radio?
Regarding my first point in the previous post:
If I have a 100-song playlist and I load all 100 songs into a taste profile, then generate a playlist using catalog-radio and specifying that taste profile, the results are unacceptable. The song selection is all over the place.<<
That observation is not correct.
I just tried again -- 100 song taste profile and generated a 2000-song playlist. Adventurousness =1, Variety =1 and Distribution = wandering. The songs returned by echonest seem to be appropriate choices based on the original 100 songs.
↧