search

wenwenyu / PICK-pytorch

Code for the paper "PICK: Processing Key Information Extraction from Documents using Improved Graph Learning-Convolutional Networks" (ICPR 2020)
https://arxiv.org/abs/2004.07464
MIT License
556 stars 193 forks source link

How to evaluate on SROIE2019? #18

Closed tengerye closed 4 years ago

tengerye commented 4 years ago

Hi, thank you for your great job.

May I ask two questions please?

  1. Did you use external data for SROIE 2019 competition?
  2. How did you evaluate on SROIE dataset?
wenwenyu commented 4 years ago

  1. External data are not used but we build a lexicon for post-processing

  2. At the training phase, we use span-based f1 metrics for logging and monitoring training process and saved bast top-k checkpoint. At the testing phase, we predict results and reconstruct to the final entity with constraint rules post-processing to correct the results then submit to the official evaluation website whose metric is entity-based f1.

n0ct4li commented 3 years ago

@wenwenyu Can you please give advices on your method for post processing for SROIE? Especially when you have multiple texts for same entity.