search

MusicConnectionMachine / Relationships

GNU Affero General Public License v3.0
9 stars 1 forks source link

Only 10 (real) artists have events? #78

Open chaoran-chen opened 7 years ago

chaoran-chen commented 7 years ago

Hey. Are you still extracting events and populating the database? I'm asking because I've been just looking into the mcmprod database and am wondering that there are already 128129 events but only 197 events belong to an artist who has a name and only 10 artists with a name have at least one event.

-- Returns 10
select count(*)
from artists a
where exists(
          select *
          from events e
          where a."entityId" = e."entityId"
      )
and a.name is not null;

Is it how it should be right now? Should I just wait for some days and there will be more useful events?

(@ansjin)

ansjin commented 7 years ago

Hopefully fixed here https://github.com/MusicConnectionMachine/Relationships/pull/79 but testing need to be done!

Sandr0x00 commented 7 years ago

Some further Information to this matter: We only link stuff when we run Wikipedia Pages. And we only can link if the Name in the Wiki-Page is exactly the name in the artists table (since we have no pseudonyms, and the code does not know if it's the same person otherwise)

In the future (makes only sense if we have pseudonyms), we additionally could run some OpenIE, Ollie, .... on the extracted date-events, gather some more information about subjects and objects and then try to link the given subjects and objects. This is a big thing, since we may get multiple subjects and objects and have to think of some magic on how to determine which is the correct one to link to. Regarding the deadline today and the status of the db, image this will and can not be done right now.

We hopefully will link some more artists with #79, but maybe not enough, or the ones for you to show @chaoran-chen