IR2 Project 2018
Papers:
-
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints - https://arxiv.org/pdf/1809.01215
-Topic Aware Neural Response Generation - http://www.aaai.org/ocs/index.php/AAAI/AAAI17/paper/download/14563/14260
-Towards Implicit Content-Introducing for Generative Short Text Conversation Systems - http://www.aclweb.org/anthology/D17-1233 -
The code for the implemented seq2seq model can be found in the file: ParlAi/parlai/agents/seq2seq_custom/seq2seq_custom.py
The PMI model is created when initializing the model (_init_model function). This is only done once, after that this is skipped. The training of the two models is done in the function train_step
The run the training process when in ParlAI folder:
sh twitter.sh