Closed Allan-waker closed 8 months ago
Thanks for your comments. For an intuitive motivation for the TCE module, please refer to our paper here: https://aclanthology.org/2023.findings-emnlp.807/.
Additionally, there was a bug in the previous version of the release. The TCE module would get trained if the hyperparameters are correctly set up (once again please refer to our paper)
What is the function of TCE module? Is it effective? I feel like the correct number of triples is not predicted well after training? I'm very much looking forward to your answer.
This is the number of triples predicted by the training process and the correct number of triples. Why are the predicted triples all between 1 and 2?