search

nanguoshun / LSR

Pytorch Implementation of our ACL 2020 Paper "Reasoning with Latent Structure Refinement for Document-Level Relation Extraction"
127 stars 22 forks source link

Some question about the Intra-F1 and Inter-F1 test #6

Closed xwjim closed 4 years ago

xwjim commented 4 years ago

In the paper, it said that the intra-f1 and inter-f1 are computed based on re-trained. How do you deal with the dataset when training intra-f1 and inter-f1. Whether you change the sents and vertexSet in the dataset

nanguoshun commented 4 years ago

hi @xwjim. We designed a rule to filter the intra- and inter-relation pairs. We didn't make any changes to the dataset. As indicated in our paper, an entity pair requires inter-sentence reasoning if the two entities from the same document have no mentions in the same sentence.

xwjim commented 4 years ago

hi @xwjim. We designed a rule to filter the intra- and inter-relation pairs. We didn't make any changes to the dataset. As indicated in our paper, an entity pair requires inter-sentence reasoning if the two entities from the same document have no mentions in the same sentence.

Do you use the evience of the relational entity pair directly or classify by the mention position to classify the entity pairs for inter and intra

nanguoshun commented 4 years ago

To identity the inter and intra-relations, we refer to the evidence indices for each entity/mentions given in the "vertexSet" field of DoRED.

nanguoshun commented 4 years ago

I will close this issue if there are no further comments