Closed kizunasunhy closed 1 year ago
Sure. This method is adopted by [1]. [1] Denoising relation extraction from document-level distant supervision. Xiao el al.
We applied the leviated markers to generate the span-pair concept in the input rather than the final layer.
Just a little question. Why do we need to pack according to subjects? Is it possible to pack all the entities with solid markers and concatenate all the possible span pairs at the end of one instance? Is it because of nested NER? Thank you.