datasenseiaryan / ts3000_thechatbot Goto Github PK

View Code? Open in Web Editor NEW

Its a social networking chat-bot trained on Reddit dataset . It supports open bounded queries developed on the concept of Neural Machine Translation. Beware of its being sarcastic just like its creator :stuck_out_tongue_closed_eyes: BDW it uses Pytorch framework and Python3.

License: MIT License

Jupyter Notebook 22.85% Python 77.15%

ts3000_thechatbot's Introduction

TS3000_TheChatBot

Tony Stark 3000 - The Chat Bot Its a very basic level conversational AI.

Its a social networking chat-bot trained on Reddit dataset . It supports open bounded queries developed on the concept of Neural Machine Translation. Beware of its being sarcastic like its creator 😝 BDW it uses Pytorch framework and Python3

Data Preprocessing :

Downoad Reddit Data From Here

Follow These :

Put Downloaded data in Data_Preprocessing directory .
Unzip the .bz2 file bzip2 -dk filename.bz2
- *Install Bzip2 on Ubuntu - sudo apt-get update - sudo apt-get install bzip2 - Useful Links : - Installing Bzip2 - Unzipping Error
Run createDB.py

python3 createDB.py

This will create Database from Raw JSON text file which you unzipped earlier.
Run createCORPUS.py

python3 createCORPUS.py

This will create corpus .For example I created 2011-08small.txt
Move this created corpus to Data directory .

Trainining Model :

Start training model using this command :

python3 main.py -tr data/2013-09small.txt -l -lr 0.0001 -it 50000 -b 64 -p 500 -s 1000
To resume training from last where yiu left :

python3 main.py -tr data/2013-09small.txt -l save/model/2013-09small/1-1_512/3000_backup_bidir_model.tar -lr 0.0001 -it 50000 -b 64 -p 500 -s 1000