Closed vahidsj closed 2 years ago
Basically, this model can do ATE and APC, but do not use BERT-SPC when the experimental results of ATE are needed. Sorry that this repo can not prepare your custom dataset, you have to construct a dataset for joint ATE-APC training according to the format of the provided datasets. Then choose a suitable pretrained-BERT to initialize and train the model.
Hi,
based on the results, described in the paper, this model works better for both ATE and APC, compared to the LCF-BERT.
I would like to know if it's possible to use this model for ATE for a new dataset. We have a gold dataset for cellphone and we want to see how this model performs. Do you have any idea how to use the trained model for new data points?
Best, Vahid