Why use multiple sigmoids for relation classifier?

lavis-nlp / spert

PyTorch code for SpERT: Span-based Entity and Relation Transformer

MIT License

685 stars 146 forks source link

Why use multiple sigmoids for relation classifier? #47

Closed syuoni closed 3 years ago

syuoni commented 3 years ago

Hi, May I ask why you use multiple sigmoids for relation classifier? Why not use a softmax layer as for the span classifier?

I have checked that there are no multiple relations between two given entities in all three datasets (only a very few duplicated relations in SciERC).

Best, Enwei

markus-eberts commented 3 years ago

Hi, this question was already answered in issue #10.