Code for paper

Yubo Xie, Ekaterina Svikhnushina, and Pearl Pu. A Multi-Turn Emotionally Engaging Dialog Model. IUI 2020 Workshop on User-Aware Conversational Agents. PDF Link.

The project was developed using the following packages:

• TensorFlow 1.12.0
• NumPy 1.15.4
• spaCy 2.0.18

• model.py: implementation of the model;
• attention.py: implementation of the hierarchical attention mechanism;
• attn_wrapper.py: wrap the attention mechanism into the RNN cells;
• train.py: train the model on the training set;
• validate.py: evaluate the model on the validation set (choosing hyperparameters);
• predict.py: evaluate the model on the test set (predict the responses).

The training data (tokenized) can be found here.

See the LICENSE file in the root repo folder for more details.