#metabrainz

/

      • d4rkie has quit
      • 2020-11-11 31659, 2020

      • D4RK-PH0ENiX joined the channel
      • 2020-11-11 31628, 2020

      • Sophist-UK has quit
      • 2020-11-11 31614, 2020

      • v6lur_ has quit
      • 2020-11-11 31616, 2020

      • ephemer0l_ has quit
      • 2020-11-11 31659, 2020

      • sumedh joined the channel
      • 2020-11-11 31613, 2020

      • sumedh has quit
      • 2020-11-11 31654, 2020

      • MajorLurker has quit
      • 2020-11-11 31615, 2020

      • ephemer0l joined the channel
      • 2020-11-11 31610, 2020

      • sumedh joined the channel
      • 2020-11-11 31639, 2020

      • supersandro2000 joined the channel
      • 2020-11-11 31616, 2020

      • sumedh has quit
      • 2020-11-11 31654, 2020

      • sumedh joined the channel
      • 2020-11-11 31632, 2020

      • prabhav3257 joined the channel
      • 2020-11-11 31614, 2020

      • prabhav3257 has quit
      • 2020-11-11 31605, 2020

      • prabhav3257 joined the channel
      • 2020-11-11 31641, 2020

      • adhi001 joined the channel
      • 2020-11-11 31604, 2020

      • prabhav3257 has quit
      • 2020-11-11 31610, 2020

      • sumedh has quit
      • 2020-11-11 31626, 2020

      • niceplace joined the channel
      • 2020-11-11 31634, 2020

      • supersandro2000 has quit
      • 2020-11-11 31652, 2020

      • supersandro2000 joined the channel
      • 2020-11-11 31633, 2020

      • sumedh joined the channel
      • 2020-11-11 31638, 2020

      • D4RK-PH0ENiX has quit
      • 2020-11-11 31626, 2020

      • D4RK-PH0ENiX joined the channel
      • 2020-11-11 31622, 2020

      • v6lur_ joined the channel
      • 2020-11-11 31601, 2020

      • Gazooo79494 has quit
      • 2020-11-11 31645, 2020

      • Gazooo79494 joined the channel
      • 2020-11-11 31634, 2020

      • leonh has quit
      • 2020-11-11 31656, 2020

      • leonh joined the channel
      • 2020-11-11 31625, 2020

      • v6lur_ has quit
      • 2020-11-11 31606, 2020

      • v6lur_ joined the channel
      • 2020-11-11 31635, 2020

      • BrainzGit
        [musicbrainz-server] reosarevok opened pull request #1785 (master…MBS-8438): MBS-8438: Only return event once in find_by methods https://github.com/metabrainz/musicbrainz-server/…
      • 2020-11-11 31637, 2020

      • BrainzBot
        MBS-8438: The same event is displayed twice on an artist's /events tab if they have multiple roles on it https://tickets.metabrainz.org/browse/MBS-8438
      • 2020-11-11 31615, 2020

      • reosarevok
        Hmm. On one hand, I wish all bugs were fixed in 15 min
      • 2020-11-11 31628, 2020

      • reosarevok
        On the other hand, how would I eat if all bugs were fixed in 15 min, I guess :p
      • 2020-11-11 31607, 2020

      • Mr_Monkey
        pristine___: Hi! Regarding your comment on 1149, is there a part of the test missing? I don't see anything related to testing the HTML elements.
      • 2020-11-11 31607, 2020

      • Mr_Monkey
      • 2020-11-11 31645, 2020

      • pristine___
        Mr_Monkey: yeah, right. I wanted to ask if that's how `updateFeedback` should be used, because I get an error if I spy on it like that
      • 2020-11-11 31657, 2020

      • pristine___
        Haven't included tbe html elements yet
      • 2020-11-11 31647, 2020

      • ruaok
        alastairp: I was pondering the name ACRP.... How about "canonical recordings"?
      • 2020-11-11 31608, 2020

      • Mr_Monkey
        I see. Well, that does look OK to me at first glance. What's the error?
      • 2020-11-11 31615, 2020

      • alastairp
        ruaok: remind me what happens if there are many matches for this exact artist credit/recording?
      • 2020-11-11 31651, 2020

      • alastairp
        I'm unsure about using "canonical" if it skips valid results
      • 2020-11-11 31621, 2020

      • Mr_Monkey
        I ran the test, and I do see the error. Let me look at it quickly
      • 2020-11-11 31639, 2020

      • ruaok
        let me explain the process... first all the releases in MB, except VA are sorted by digital release, release year, with the goal of matches settling on "things that reasonably came from a digital release".
      • 2020-11-11 31604, 2020

      • ruaok
        then that list is fetched and the artist-credit, recordings "pairs" are recorded in an in memory dict.
      • 2020-11-11 31648, 2020

      • ruaok
        and the first occurance of any one pair is attributed to its source release and the recording MBID from that release is chosen as the "canonical one"
      • 2020-11-11 31647, 2020

      • ruaok
        if we add VA and fix the releases without a country TODOs then all text combos of artist credits and recordings will be in the mapping.
      • 2020-11-11 31656, 2020

      • alastairp
        yeah, right. so I'm asking about the non-common cases, where there are valid cases where there are mroe than 2 valid recordings for the same (ac, r name) pair
      • 2020-11-11 31629, 2020

      • ruaok
        all but the first are discarded.
      • 2020-11-11 31644, 2020

      • ruaok
        its a text lookup, not an MBID lookup.
      • 2020-11-11 31605, 2020

      • ruaok
        maybe canonical is too specific. any other ideas for better names?
      • 2020-11-11 31618, 2020

      • alastairp
        my first gut feel is that "canonical" makes it sound like it's pretty damn near perfect, and I think there are a lot of edge cases that it skips
      • 2020-11-11 31636, 2020

      • alastairp
        sorry, this isn't super helpful in proposing an alternative, I just wanted to comment on it
      • 2020-11-11 31647, 2020

      • ruaok
        wishy washy recording mapping??
      • 2020-11-11 31654, 2020

      • alastairp
        that's a lot clearer :)
      • 2020-11-11 31609, 2020

      • alastairp
        honestly, artist credit recording mapping
      • 2020-11-11 31611, 2020

      • ruaok
        tech is easy, naming is hard.
      • 2020-11-11 31617, 2020

      • alastairp
        is clearer that the P
      • 2020-11-11 31629, 2020

      • ruaok
        oh heh. sure.
      • 2020-11-11 31643, 2020

      • ruaok
        I was blind to that since that was never meant to be a mapping in the first place. lol.
      • 2020-11-11 31615, 2020

      • ruaok
        ok, I won't fix the name just yet. I'll keep going until we're sure that it serves our needs. then I'll do a refactor/cleanup on it.
      • 2020-11-11 31621, 2020

      • alastairp
        sounds fine to me
      • 2020-11-11 31626, 2020

      • alastairp
        Mr_Monkey: pristine___: hello!
      • 2020-11-11 31641, 2020

      • Mr_Monkey
        Ahoy ! Let's talk feedback :)
      • 2020-11-11 31631, 2020

      • pristine___
        Hi!
      • 2020-11-11 31635, 2020

      • alastairp
        pristine___: I had a look at your PR, it looks mostly good
      • 2020-11-11 31658, 2020

      • alastairp
        however, last week Mr_Monkey and I had a discussion at the office, and we think that there might have been some confusion about the purpose of this feature and what it needs to do
      • 2020-11-11 31620, 2020

      • alastairp
        it's possible that we didn't clearly specify what we were doing - first up I wanted to ask if there was a ticket describing what we're doing
      • 2020-11-11 31624, 2020

      • Mr_Monkey
        I was definitely confused.
      • 2020-11-11 31653, 2020

      • pristine___
        No ticket
      • 2020-11-11 31635, 2020

      • alastairp
        I think that's a good thing to focus on first. This is similar to what we were talking about yesterday - we didn't have a clear "what do we want to do" aim, it was more of a "hey let's do some collaborative filtering"
      • 2020-11-11 31602, 2020

      • alastairp
        let me make a comment about my understanding of the task first, and then we can continue from there
      • 2020-11-11 31640, 2020

      • alastairp
        we currently have feedback, a way for users to say that they like or dislike something that they have listened to.
      • 2020-11-11 31641, 2020

      • pristine___
        alastairp: I thought we are clear on purpose because we had a discussion about feedback stuff when we were writing the schema
      • 2020-11-11 31602, 2020

      • pristine___
        And talking about the 5 feedback options
      • 2020-11-11 31603, 2020

      • alastairp
        This current implementation is problematic in a few ways: it's mapped directly to a msid, and it only has 2 types of feedback, positive and negative
      • 2020-11-11 31615, 2020

      • pristine___
        It is mapped to mbid
      • 2020-11-11 31630, 2020

      • alastairp
        so, we decided to redo feedback, adding in 4 + 1 types of feedback, and map it to an mbid
      • 2020-11-11 31631, 2020

      • pristine___
        Wait, are we taking about recommendation feedback?
      • 2020-11-11 31652, 2020

      • alastairp
        pristine___: right. and this brings up the main difference and point of confusion that I think we're seeing
      • 2020-11-11 31602, 2020

      • pristine___
        Because there is something called listen feedback too
      • 2020-11-11 31606, 2020

      • ruaok
        that was beautiful. lol
      • 2020-11-11 31607, 2020

      • pristine___
        Ooo. Right
      • 2020-11-11 31615, 2020

      • alastairp
        I get the impression that your work has been "feedback to recommendations"
      • 2020-11-11 31621, 2020

      • pristine___
        Yes
      • 2020-11-11 31631, 2020

      • alastairp
        and my proposal was that this shoud be an improvement to "feedback"
      • 2020-11-11 31658, 2020

      • alastairp
        that is, we're not sure that it's useful to have general "recommendation feedback"
      • 2020-11-11 31658, 2020

      • pristine___
        Okay, so I think we can roll in shivam-kapila too.
      • 2020-11-11 31606, 2020

      • pristine___
        Hmm
      • 2020-11-11 31616, 2020

      • alastairp
        this comes back to the initial comment about if we have a ticket describing what we want
      • 2020-11-11 31651, 2020

      • alastairp
        my opinion: I don't think that we should be asking people if they love/like/don'tlike/hate a track _in the context of a recommendation_
      • 2020-11-11 31613, 2020

      • alastairp
        we should just ask people in general "do you like/love/don'tlike/hate this track"
      • 2020-11-11 31625, 2020

      • alastairp
        now, bad recommendation - that's a slightly different case
      • 2020-11-11 31633, 2020

      • supersandro2000 has quit
      • 2020-11-11 31650, 2020

      • supersandro2000 joined the channel
      • 2020-11-11 31653, 2020

      • alastairp
        (I'll take a pause here in case you have any comments)
      • 2020-11-11 31654, 2020

      • SothoTalKer has quit
      • 2020-11-11 31611, 2020

      • pristine___
        No comments
      • 2020-11-11 31621, 2020

      • alastairp
        ok, cool. continuing then
      • 2020-11-11 31625, 2020

      • alastairp
        you're right - we had a discussion about how it would work, and you sent the schema. I didn't notice that you had named the table recommendation_feedback, I think that was the first indication that we might have been understanding different things
      • 2020-11-11 31603, 2020

      • pristine___
        I named in light of `bad_recommendation`
      • 2020-11-11 31608, 2020

      • pristine___
        Right
      • 2020-11-11 31609, 2020

      • SothoTalKer joined the channel
      • 2020-11-11 31633, 2020

      • alastairp
        we suggested to add the "bad recommendation" feedback too. This is because we were thinking in general about what kind of feedback a person might want to give in all cases
      • 2020-11-11 31647, 2020

      • alastairp
        consider "bad recommendation". It doesn't actually say "I like this track" or "I hate this track"
      • 2020-11-11 31656, 2020

      • alastairp
        it's just saying "in this context, it's not a great suggestion"
      • 2020-11-11 31659, 2020

      • alastairp
        so I think it's true that we're confusing a few things, and we could have simplified it and reduced some confusion by splitting it into 2 parts
      • 2020-11-11 31610, 2020

      • alastairp
        but that's OK, we can continue from where we are
      • 2020-11-11 31617, 2020

      • alastairp
        Mr_Monkey: want to chime in about ordering?
      • 2020-11-11 31624, 2020

      • pristine___
        What I think is
      • 2020-11-11 31612, 2020

      • Mr_Monkey
        Didn't quite understand what you mean by "ordering"
      • 2020-11-11 31617, 2020

      • pristine___
        If we want to continue from where we are and if we want to keep the feedback stuff consistent, we can reuse the rec feedback component for listens feedback and just remove the `bad_recommendation` option
      • 2020-11-11 31652, 2020

      • Mr_Monkey
        Close to that ! I think we'll want to *separate* the bad_recommendation feedback from the rest of the feedback.
      • 2020-11-11 31609, 2020

      • pristine___
        Yup
      • 2020-11-11 31611, 2020

      • Mr_Monkey
        And show it only on recommendation cards
      • 2020-11-11 31615, 2020

      • alastairp
        Mr_Monkey: sorry, ordering of the buttons
      • 2020-11-11 31634, 2020

      • Mr_Monkey
        While we show the "emoticons" feedback on all listens and rec cards.
      • 2020-11-11 31647, 2020

      • pristine___
        That Mr_Monkey
      • 2020-11-11 31615, 2020

      • alastairp
        cool
      • 2020-11-11 31617, 2020

      • Mr_Monkey
        I think, since the emoticons component will be consistently visible, we'll want to show it to the right (since that part of the cards in right-aligned). That way it should always be in the same position.
      • 2020-11-11 31618, 2020

      • Mr_Monkey
        That means we would show another sort of rec-only feedback component to the left of it.
      • 2020-11-11 31654, 2020

      • alastairp
        so I think that _for now_ it's OK to keep the 2 types of feedback - we need to do more work in the msid mapping to be able to use this feedback component on the main listen view, right?
      • 2020-11-11 31655, 2020

      • pristine___ tries to imagine
      • 2020-11-11 31640, 2020

      • Mr_Monkey
        Or, alternatively, find a way to modify the existing emoticons dropdown to make a separation between bad_rec and the rest.
      • 2020-11-11 31658, 2020

      • pristine___
        Yeah, if listens are mapped to msid mapping, we will have an mbid for each recording. It will also require feedback schema update ig
      • 2020-11-11 31628, 2020

      • pristine___
        Mr_Monkey: just to confirm, in a separate PR? #1149 has grown bigggg
      • 2020-11-11 31631, 2020

      • pristine___
        :p
      • 2020-11-11 31636, 2020

      • Mr_Monkey
        Yes, separate PR :)
      • 2020-11-11 31629, 2020

      • Mr_Monkey
        Thankfully, I believe none of the work you did will be lost, just modified.
      • 2020-11-11 31643, 2020

      • alastairp
        yeah, agreed. this isn't a huge change from what is already there
      • 2020-11-11 31644, 2020

      • pristine___
        Yup!
      • 2020-11-11 31650, 2020

      • Mr_Monkey
        Phew !
      • 2020-11-11 31607, 2020

      • alastairp
        I suggest that we modify the endpoints and database tables to remove the references to "recommendation feedback" where possible
      • 2020-11-11 31603, 2020

      • pristine___
        Hmm. I think we can plan a road map here, as in what to do first and next and so on
      • 2020-11-11 31629, 2020

      • alastairp
        can you make a list and show it to us later today?
      • 2020-11-11 31631, 2020

      • Mr_Monkey
        I propose that I make a couple of prototypes of our options for separating the bad_rec feedback, while you're busy finishing 1149 and modifying as alastairp suggests
      • 2020-11-11 31643, 2020

      • pristine___
        Cool
      • 2020-11-11 31643, 2020

      • Mr_Monkey
        (Just visual ones)
      • 2020-11-11 31645, 2020

      • pristine___
        Cool cool
      • 2020-11-11 31649, 2020

      • alastairp
        thanks pristine___ !
      • 2020-11-11 31605, 2020

      • pristine___
        :)
      • 2020-11-11 31607, 2020

      • alastairp
        (there's one more thing that I need to comment about, but we can't do anything about it until we make playlists)
      • 2020-11-11 31632, 2020

      • pristine___
        ?
      • 2020-11-11 31633, 2020

      • alastairp
        the problem is that "this is a bad recommendation" needs a context. just saying "recording mbid is a bad recommendation" isn't enough information for us to do it
      • 2020-11-11 31607, 2020

      • Mr_Monkey
        -> "this is a bad recommendation" + "in the context of playlist #12345"
      • 2020-11-11 31608, 2020

      • alastairp
        so once we are able to store recommendation lists in a playlist in listenbrainz, we can say "this recording is a bad recommendation in <this specific playlist>"
      • 2020-11-11 31615, 2020

      • pristine___
        Right. I was thinking to add some context to it for users to understand what exactly this option is in the rec info page
      • 2020-11-11 31649, 2020

      • Mr_Monkey
        I think that will be useful