thank you for your amazing work on including entity information in the language model.
I'm considering computing an entity similarity using entity vectors extracted from LUKE's entity embedding's last hidden state.
The similarity is between an entity in a sentence and LUKE's entity vocab.
Do you think this makes sense? I wonder if you could shed light on this kind of usage of LUKE.
Thank you in advance! :)
Hi,
thank you for your amazing work on including entity information in the language model.
I'm considering computing an entity similarity using entity vectors extracted from LUKE's entity embedding's last hidden state. The similarity is between an entity in a sentence and LUKE's entity vocab.
Do you think this makes sense? I wonder if you could shed light on this kind of usage of LUKE. Thank you in advance! :)