ChatBotWithSCST
What is it?
This is a simple chat bot for entertainment.
The bot learned from Cornell Movie-Dialogs Corpus dataset, using seq2seq model.
The seq2seq model is fine-tuned by REINFORCE algorithm (and self-critical sequence training based on it).
Chat bot is implemented using python-telegram-bot
Dataset
This project used Cornell Movie-dialogs Corpus (Cornell University) as a learning dataset.
The dataset can be downloaded by running ./data/download_data.sh.
Requirements
Environment
This bot can be both trained and used on Google Colab through notebooks.
Configurations
- Make directory
./.config - Make file
./.config/telegram_conf.ini
Example format of telegram_conf.ini:
[TELEGRAM]
APIToken = HERE_IS_YOUR_API_TOKEN [required]
[BOT]
CallName = CALL_NAME_OF_BOT
BotName = REAL_NAME_OF_BOT
...