This repository contains a generative chatbot that takes a conversation history as input and generates a valid response. The dataset used for this task is Daily Dialogues.
The chatbot is designed with the assumption that the first speaker is the user and the second one is the agent. All turns until the agent's last turn are taken as conversation history, and the chatbot predicts the last agent's message. The chatbot uses a transformers model for response generation and a Gradio interface for interaction. This project also includes a bonus feature that leverages the emotion labels provided in the dataset to improve response generation by making the chatbot emotion-aware.
To set up the environment for this project, ensure Anaconda or Miniconda is installed. Then, you can create the Conda environment and install the dependencies with:
conda env create -f environment.ymlActivate the environment with:
conda activate generative_chatbotTo start the Gradio interface, run:
. ./test.sh