README.md

DRL4DG : Deep Reinforcement Learning for Dialogue Generation using SEQ2SEQ model

Goal :

The goal of this repository is to develop a Dialogue Agent which generates coherent and interesting dialogues based on Seq2Seq Model and Deep Reinforcement Learning approach. We also would like to integrate some emotion management to make the agent empathetic.

Dataset :

The dataset used is the DailyDialog dataset that you can download here.

Pre-requisites :

The code is deveopped using Anaconda with a custom evironment composed of the following main packages :

• python 3.7.9
• pytorch 1.5.1
• cudatoolkit 9.2

Resources :

This work is based on the Deep Reinforcement Learning for dialogue generation paper using PyTorch. In order to get some base notions, we started our work with this PyTorch official Bot tutorial and this repository.