Conversation

Notices

1. Embed this notice
   Kate Nyhan (kdnyhan@social.esmarconf.org)'s status on Thursday, 05-Dec-2024 10:33:51 JST Kate Nyhan

   in reply to

   • mshoberecht

   @mshoberecht
   Hoorah, we are saving money and time and "overall" the quality is 74% as good as it used to be!

   • Embed this notice
     Kate Nyhan (kdnyhan@social.esmarconf.org)'s status on Thursday, 05-Dec-2024 10:33:53 JST Kate Nyhan

     in reply to

     • mshoberecht

     @mshoberecht
     And -- look, I'm just watching from the outside, maybe my guesses are wrong -- but we all know that named entity recognition for things like diseases can perform quite well. So if the disease terms are presumably being correctly applied to PubMed abstracts by MTIX, that suggests that the *non-disease* subject headings are being applied incorrectly, or missed when they should be applied, at a rate that is even worse than the 66% headline "other descriptors" score.

   • Embed this notice
     Kate Nyhan (kdnyhan@social.esmarconf.org)'s status on Thursday, 05-Dec-2024 10:34:16 JST Kate Nyhan

     • mshoberecht

     @mshoberecht
     I am focusing on the "pther descriptors" category, not because I'm cherry-picking the lowest, but because other descriptors than check tags and publication types means -- the MeSH terms that #medlibs use for searching -- the headings like diseases and outcomes, populations that aren't checktags, research methods that aren't publication types....