Junki Ohmura

Podcastでも解説しました． https://anchor.fm/lnlp-ninja/episodes/ep50-ICLR-ELECTRA-Pre-training-Text-Encoders-as-Discriminators-Rather-Than-Generators-ebgu2f

https://github.com/arXivTimes/arXivTimes/issues/645

Podcastでも解説しました https://anchor.fm/lnlp-ninja/episodes/ep46-FreeLB-Enhanced-Adversarial-Training-for-Language-Understanding-e9uauc

Podcastでも解説しました https://anchor.fm/lnlp-ninja/episodes/ep47-ALBERT-A-Lite-BERT-for-Self-supervised-Learning-of-Language-Representations-ea4jv4

Podcastでも紹介しました． https://anchor.fm/lnlp-ninja/episodes/ep35-Modeling-Semantic-Relationship-in-Multi-turn-Conversations-with-Hierarchical-Latent-Variables-e4fss7

Poscastでも解説しました． https://anchor.fm/lnlp-ninja/episodes/ep34-Do-Neural-Dialog-Systems-Use-the-Conversation-History-Effectively-e4958g

Podcastでも解説しました． https://anchor.fm/lnlp-ninja/episodes/ep45-Episodic-Memory-in-Lifelong-Language-Learning-e98uup

@oneTaken Hi. Thank you for your comments. That will be nice to debug. I am also still facing the low performance problem. I am exactly checking now. If you find...

I am checking the code by removing p2 layer (only predicts the answer of beginning)

Last layer should predict the index of word instead of char level index. https://github.com/jojonki/BiDAF/commit/91aa342da6ff6e807e19e3bd60ff42101426da5d#diff-8f329902bf6ef6e1af03f01d0b9633e2