Forum » Feedback and Ideas

Automated & manual Artist disambiguation: How to drastically improve this site,…

 
  • WichitaQ said:
    iTranscendence said:
    An artist ID would work for the databasing, but it doesn't help someone who is using google, or the internal search engine at last.fm which is why you just put the (2), (3) etc after the names, then when you search them it brings up the whole list, and could have their style of music, to help the user find it quicker.


    artist disambiguation page?
    i don't see why adding (2), (3) and similar would help, how would we still know which band we're looking for without the additional info about them. if we were to use a disambiguation page we would have a list of all the artists sharing a name and a quick description about them. much easier solution, IMHO.

    edit: and yes, artist ID would still be used in a database, but they would have their name, i surely would want to type rf63df48h9-6ghze23 to search for a band :D
    different artists with a same name can indeed share a name, but they can't share an ID.


    It wouldn't which is why i suggested a 'disambiguation' page. May be you mis read me but that is exactly what I was proposing.

  • Bara_Megami said:
    I agree with the original poster.

    For me, last.fm is about the statistics. I couldn't care less if the radio is charging money, since I don't use it regularly (I do feel for the people who do though - if last.fm started charging for stats, I'd be pissed). So needless to say, I feel that the slideshow radio is totally useless (I still have some on my free trial), and the staff could be working on making the stats more accurate. You know, instead of making a pretty slideshow app.

    We need merging and automatic redirecting of incorrectly spelled artist names and separation of same named artists. Otherwise, it looks like I'm listening to the Metal bands "Aria" and "Sakura" instead of the female J-hiphop artists with the same name.

    I don't mean to be cruel, but...come on. The automatic corrector needs to be a priority.


    The issue isn't the auto correcter so much as for some reason beyond my comprehension they did not devise a way for the audioscrobbler to separate bands with the same name. This can be solved fairly simply with the 'band id' for internal use and the discogs technique for the end user.

    • gwalla sa...
    • Användare
    • 12 maj 2009, 06:30
    Disambiguation pages work great for Wikipedia...but Wikipedia is not dependent on ID3 tags for its information, and all of Wikipedia's information is meant to be used directly by humans interacting with the site. Last.FM, on the other hand, collects most of its information automatically (through passive scrobbling) and uses it to fuel computer algorithms, without any direct user input.

    Many, probably most, tracks in the Last.FM database were not submitted directly by a label or artist. All Last.FM knows about them is that somebody played a song with one string of characters in the artist tag, another string of characters in the title tag, and maybe another string of characters in the album tag.

  • gwalla said:
    Disambiguation pages work great for Wikipedia...but Wikipedia is not dependent on ID3 tags for its information, and all of Wikipedia's information is meant to be used directly by humans interacting with the site. Last.FM, on the other hand, collects most of its information automatically (through passive scrobbling) and uses it to fuel computer algorithms, without any direct user input.

    Many, probably most, tracks in the Last.FM database were not submitted directly by a label or artist. All Last.FM knows about them is that somebody played a song with one string of characters in the artist tag, another string of characters in the title tag, and maybe another string of characters in the album tag.


    I'm not saying one model or the other is specifically superior. I'm saying that the site over all could benifit from a hybrid model of the two working in concert.

    • maz35 sa...
    • Användare
    • 12 maj 2009, 13:44
    iTranscendence said:
    gwalla said:
    Disambiguation pages work great for Wikipedia...but Wikipedia is not dependent on ID3 tags for its information, and all of Wikipedia's information is meant to be used directly by humans interacting with the site. Last.FM, on the other hand, collects most of its information automatically (through passive scrobbling) and uses it to fuel computer algorithms, without any direct user input.

    Many, probably most, tracks in the Last.FM database were not submitted directly by a label or artist. All Last.FM knows about them is that somebody played a song with one string of characters in the artist tag, another string of characters in the title tag, and maybe another string of characters in the album tag.


    I'm not saying one model or the other is specifically superior. I'm saying that the site over all could benifit from a hybrid model of the two working in concert.


    I think they are actually working on such a thing, however its taking them forever because it is nowhere near as easy as your making it sound.

  • maz35 said:
    iTranscendence said:
    gwalla said:
    Disambiguation pages work great for Wikipedia...but Wikipedia is not dependent on ID3 tags for its information, and all of Wikipedia's information is meant to be used directly by humans interacting with the site. Last.FM, on the other hand, collects most of its information automatically (through passive scrobbling) and uses it to fuel computer algorithms, without any direct user input.

    Many, probably most, tracks in the Last.FM database were not submitted directly by a label or artist. All Last.FM knows about them is that somebody played a song with one string of characters in the artist tag, another string of characters in the title tag, and maybe another string of characters in the album tag.


    I'm not saying one model or the other is specifically superior. I'm saying that the site over all could benifit from a hybrid model of the two working in concert.


    I think they are actually working on such a thing, however its taking them forever because it is nowhere near as easy as your making it sound.


    No from a web design/ web development perspective it's not a walk in the park, but the best path isn't always the easiest one.

  • Hi,

    I've just read through the thread, and although the same artist thing doesn't bother me (I don't think I've ever been confused about which artist is meant), it all boils down to the original data source. As has been said, if you want two bands with the same name to be distinguished between, then they need a unique ID (how that is display: "Tom Jones", "Tom Jones (2)" is a matter of taste and could be last.fm user preference based on a few different options (numbering, by year of first release, country, etc.)). The Unique IDs of every track, artist, etc., needs to be developed first, then the rest should fall into place quite easily. I currently use MusicBrainz as it is quick and easy and hits probably 80% of tracks; for the other 20%, there are very few I care enough about to go to discogs/freedb and get the info on (yep, I'm one of those lazy leecher types). Unfortunately for MusicBrainz, it has to keep its own database...which is where we hit the main issue, every service keeps their own database.

    Imagine if you wanted to phone a directory enquiries service to get a friend's phone number, you wouldn't expect to have to phone 2-3 services, you would phone 1 of them, and they would have all the data, as it would be supplied by the same sources (i.e. the telecomms providers).

    Similarly, what needs to happen, is the major labels/distributors should be submitting full details, along with artist IDs, musician and Author IDs for each track, release IDs (like an ISBN #), and PUID (or some other open format for hashing the actual audio into a simple unique code) to a few database services like GraceNote, Freedb.org, etc. That would put Musizbrainz out of business, but the excellent Picard Tagger could just redirect to the Freedb database (after a reasonably amount of modification). Then Last.fm could display any way they wanted really...the main database services could move to not being directly accessible, but only act as primary sources for local mirroring services similarly to the way the first level DNS and NTP servers work. Thus, Last.fm would pay a license fee to freedb.org, using advertising or whatever to pay their fees, and we're all sorted.

    Obviously, this will never happen, as the last thing the major labels want to do is be "helpful"...but as it would help the artists themselves, some of the younger ones may be encouraged to help out.

    :)

  • DJMatus23 said:
    Hi,

    I've just read through the thread, and although the same artist thing doesn't bother me (I don't think I've ever been confused about which artist is meant), it all boils down to the original data source. As has been said, if you want two bands with the same name to be distinguished between, then they need a unique ID (how that is display: "Tom Jones", "Tom Jones (2)" is a matter of taste and could be last.fm user preference based on a few different options (numbering, by year of first release, country, etc.)). The Unique IDs of every track, artist, etc., needs to be developed first, then the rest should fall into place quite easily. I currently use MusicBrainz as it is quick and easy and hits probably 80% of tracks; for the other 20%, there are very few I care enough about to go to discogs/freedb and get the info on (yep, I'm one of those lazy leecher types). Unfortunately for MusicBrainz, it has to keep its own database...which is where we hit the main issue, every service keeps their own database.

    Imagine if you wanted to phone a directory enquiries service to get a friend's phone number, you wouldn't expect to have to phone 2-3 services, you would phone 1 of them, and they would have all the data, as it would be supplied by the same sources (i.e. the telecomms providers).

    Similarly, what needs to happen, is the major labels/distributors should be submitting full details, along with artist IDs, musician and Author IDs for each track, release IDs (like an ISBN #), and PUID (or some other open format for hashing the actual audio into a simple unique code) to a few database services like GraceNote, Freedb.org, etc. That would put Musizbrainz out of business, but the excellent Picard Tagger could just redirect to the Freedb database (after a reasonably amount of modification). Then Last.fm could display any way they wanted really...the main database services could move to not being directly accessible, but only act as primary sources for local mirroring services similarly to the way the first level DNS and NTP servers work. Thus, Last.fm would pay a license fee to freedb.org, using advertising or whatever to pay their fees, and we're all sorted.

    Obviously, this will never happen, as the last thing the major labels want to do is be "helpful"...but as it would help the artists themselves, some of the younger ones may be encouraged to help out.

    :)


    Now you are thinking with portals.


    • [Raderad användare] sa...
    • Användare
    • 13 maj 2009, 14:50
    The ID3 would mess things up. How could the last.fm engine know if you listen to ARTIST or ARTIST (1)? Doubt that people would start re-tagging their mp3 collections. I Know I wouldn't.

    • maz35 sa...
    • Användare
    • 13 maj 2009, 16:35
    People wont change their tags and it would make things way too complex to force them to.


    Another issue is that last.fm has scrobbles of artists so unknown that they wont even appear on wiki or discogs. Being an auto generated system last.fm has many issues the likes of wikipedia don't have to worry about.

  • JLizard79 said:
    The ID3 would mess things up. How could the last.fm engine know if you listen to ARTIST or ARTIST (1)? Doubt that people would start re-tagging their mp3 collections. I Know I wouldn't.


    Did you read the entire thread? We already discussed that the (2) (3) would be for external use of the end user, when searching for different artists, and that for the databasing each artist would have a unique "artist ID". An 'artist id' which would be tied to all the songs and albums in their catalog thus allowing the audioscrobbler to automatically detect whether you were listening to "Aes Dana" or "Aes Dana (2)".

    you = not thinking with portals.

    Redigerad av iTranscendence den 13 maj 2009, 20:40
  • maz35 said:
    People wont change their tags and it would make things way too complex to force them to.


    Another issue is that last.fm has scrobbles of artists so unknown that they wont even appear on wiki or discogs. Being an auto generated system last.fm has many issues the likes of wikipedia don't have to worry about.


    Don't be such a nay-sayer, the premise that things need to stay the way they are is simply ludicrous. Did you even read the thread?

    • [Raderad användare] sa...
    • Användare
    • 13 maj 2009, 21:55
    I like the idea but since last.fm don't allow regular users to add albums their engine won't know which songs and albums belongs to which of the bands if you scrobble bootlegs, demos and so on that aren't uploaded by the artist/ label.
    Don't take this as negativ critic, I'm just curious about what ideas you've got to solve the problems.

    • maz35 sa...
    • Användare
    • 13 maj 2009, 22:56
    iTranscendence said:
    maz35 said:
    People wont change their tags and it would make things way too complex to force them to.


    Another issue is that last.fm has scrobbles of artists so unknown that they wont even appear on wiki or discogs. Being an auto generated system last.fm has many issues the likes of wikipedia don't have to worry about.


    Don't be such a nay-sayer, the premise that things need to stay the way they are is simply ludicrous. Did you even read the thread?


    I'm not saying things need to stay as they are, I want artists to be separated eventually. I'm just stating the facts that make this a very difficult issue to fix. I'm sure the staff know how they'd arrange it on a database level its the actual separating of the data for this correctly is the problem.

    There are old blog entries on the issues facing the staff on this. The recently release auto correct system is the first step toward a better catalogue here and that still needs optimising yet.

  • JLizard79 said:
    I like the idea but since last.fm don't allow regular users to add albums their engine won't know which songs and albums belongs to which of the bands if you scrobble bootlegs, demos and so on that aren't uploaded by the artist/ label.
    Don't take this as negativ critic, I'm just curious about what ideas you've got to solve the problems.


    well until an album is officiated by an artist or label it could just go under a 'white label' section. That's something discogs has been considering since they don't currently allow any page that doesnt have a manufacturer # for that CD.

  • maz35 said:
    iTranscendence said:
    maz35 said:
    People wont change their tags and it would make things way too complex to force them to.


    Another issue is that last.fm has scrobbles of artists so unknown that they wont even appear on wiki or discogs. Being an auto generated system last.fm has many issues the likes of wikipedia don't have to worry about.


    Don't be such a nay-sayer, the premise that things need to stay the way they are is simply ludicrous. Did you even read the thread?


    I'm not saying things need to stay as they are, I want artists to be separated eventually. I'm just stating the facts that make this a very difficult issue to fix. I'm sure the staff know how they'd arrange it on a database level its the actual separating of the data for this correctly is the problem.

    There are old blog entries on the issues facing the staff on this. The recently release auto correct system is the first step toward a better catalogue here and that still needs optimising yet.


    I think you are focusing in a little too myopically on the artists sharing page issue. I was trying to bring up the broader concept of a hybrid wiki style site to help the scrobbler increase its accuracy and efficiency.

    • maz35 sa...
    • Användare
    • 14 maj 2009, 01:44
    iTranscendence said:
    maz35 said:
    iTranscendence said:
    maz35 said:
    People wont change their tags and it would make things way too complex to force them to.


    Another issue is that last.fm has scrobbles of artists so unknown that they wont even appear on wiki or discogs. Being an auto generated system last.fm has many issues the likes of wikipedia don't have to worry about.


    Don't be such a nay-sayer, the premise that things need to stay the way they are is simply ludicrous. Did you even read the thread?


    I'm not saying things need to stay as they are, I want artists to be separated eventually. I'm just stating the facts that make this a very difficult issue to fix. I'm sure the staff know how they'd arrange it on a database level its the actual separating of the data for this correctly is the problem.

    There are old blog entries on the issues facing the staff on this. The recently release auto correct system is the first step toward a better catalogue here and that still needs optimising yet.


    I think you are focusing in a little too myopically on the artists sharing page issue. I was trying to bring up the broader concept of a hybrid wiki style site to help the scrobbler increase its accuracy and efficiency.


    Now you've confused me, how is this whole discussion not entirely related to artists sharing pages? And yes I've read the whole thread. I don't understand what use a disambiguation page would be if you couldn't split the artist data up, otherwise all links will go to the same page anyway.

    As for at the scrobbler end I see suggestion for getting user input on a new unknown track when the whole idea is for it to run silently in the background scrobbling. Back when this project started the scrobblers where just simple background plugins for media players with no app of its own to ask for user input anyway. Artist wiki's didn't even exist back then. I like the fact that if I listened to a brand new artist/track they'd have an auto generated entry on last.fm almost immediately(its happened) and any ideas that suggest that they'd have to go through some approval process before they appeared on my charts concern me.

    I don't mean to be a stick in the mud about this issue and maybe I'm not understanding the suggestions properly but as much I'd love for it to be fixed tomorrow if possible it doesn't mean I wont challenge ideas on how to do it if I'm sceptical. I just think if there was a simple fix it would have been done by now.

    • Bloopy sa...
    • Forum Moderator
    • 14 maj 2009, 11:04
    I think the best way to solve it is to make the current artist pages disambiguation pages. Similar to how it is now, but with links to each artist with that name. A unique page for each artist is the easy part. It's the redirects which are tricky. By default, all scrobbles should go to the most popular artist with that name, and then specific tracks can be redirected based on MusicBrainz data and user input.

    Maybe there could be user-specific redirects, so that a user can specify which artist with that name they are listening to. If they listen to more than one artist with the same name, then they could redirect some of the tracks to the correct page. Even if two artists with the same name have a track sharing the same name, different users could have their own user-specific redirects to different pages.

  • Those are all good idea's bloopy.

    • [Raderad användare] sa...
    • Användare
    • 29 jun 2009, 00:37
    I'm really liking this idea but unfortunately last.fm has turned into a sinking ship=(
    Albums are a mess and case tags are a mess, xml using musicbrainz model without updating it. What once looked promising has now turned more and more unpromising.=(

  • Re:How to drastically improve this site: Follow the discogs.com model

    Bump for great justice.

    • [Raderad användare] sa...
    • Användare
    • 28 jul 2009, 20:15
    It a very good idea.

  • Apparently not good enough... lol

    • [Raderad användare] sa...
    • Användare
    • 29 jul 2009, 20:37
    I see another of my favourite bands (Panic Room) have had their last.fm page crapped on another band which didn't exist a few months ago.

    Once you get outside the major label mainstream, this sort of thing is becoming more and more common - it seems there are more bands that share names than don't. You get energy wasted in pissing contests between fans of the two bands over who's first in the wiki, who's photo is shown first, etc.

    Last.fm have got to get their finger out and fix this - I'm sick and tired of hearing excuses as to why it can't be done. I work in IT, I've got a lot of database design experience - I know full well that you'd be bullshitting when you say it's "impossible". If it requires a major database redesign, then so be it. You really can't afford not to.

    Even if the fix isn't perfect, and might take some sort of manual intervention to determine which 'new' tracks belong to which band, anything is preferable to the present mess. You will find fans will be prepared to spend time and energy if there's any manual disambiguation required.

    If last.fm doesn't fix this, and fix this soon, then people are going to be abandoning last.fm for rival sites.

    • [Raderad användare] sa...
    • Användare
    • 29 jul 2009, 21:03
    I see the excact same problem on spotify, but that I can take since it's so new but last.fm has been around since 2004 if I'm not misstaken.

Anonyma användare kan inte skriva inlägg. Vänligen logga in eller skapa ett konto för att göra inlägg i forumen.