You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Jonas from Riksdagen found a case of mistaken speaker attribution for one of the speeches I had matched audio and text protocols for.
Seems like it's a rare case of your speaker introduction model mistaking an utterance in a speech with a speaker introduction. You can see why the model classifies it as a speaker introduction, since the speaker says:
Herr Erlander säger nu:
But this is part of the speech itself (and not a new speaker introduction). See:
I'm not sure whether something like this warrants an issue. But since it's one of the few rare failure cases of speaker introduction I've seen I thought I'd open this issue.
The text was updated successfully, but these errors were encountered:
Jonas from Riksdagen found a case of mistaken speaker attribution for one of the speeches I had matched audio and text protocols for.
Seems like it's a rare case of your speaker introduction model mistaking an utterance in a speech with a speaker introduction. You can see why the model classifies it as a speaker introduction, since the speaker says:
But this is part of the speech itself (and not a new speaker introduction). See:
https://github.com/swerik-project/riksdagen-records/blob/main/data/1966/prot-1966--ak--030.xml#L3350-L3351
I'm not sure whether something like this warrants an issue. But since it's one of the few rare failure cases of speaker introduction I've seen I thought I'd open this issue.
The text was updated successfully, but these errors were encountered: