Any news on the pretrained entity linking model? #6665
Replies: 2 comments
-
This is not currently in the pipeline. The main issue is that it's simply hard to decide what a pretrained EL model should do. Which data should it be trained on? Which knowledge base should it use? Which type of entities should it try to resolve? Prior probabilities can vary so much between different domains/text sources/communities, that it's a huge challenge to define a model that would be generalizable. It's a little like Text categorization. We currently don't provide pretrained models, but we provide all the tools for you to be able to train your own on your specific use-case. |
Beta Was this translation helpful? Give feedback.
-
No problem. That's good to know! |
Beta Was this translation helpful? Give feedback.
-
Hi there!
Apologies if this is being discussed elsewhere. I was just wondering if there had been any developments in making the pretrained entity linking models available? See here for the last discussion I can find about them: #5156
Many thanks!
Beta Was this translation helpful? Give feedback.
All reactions