ods_help_bot

About

This project is a telegram bot trained on Open Data Science slack posts history

Structure

• chatbot.py - telegram bot for messaging
• support_model.py - question answering AI
• notebooks/know_base.ipynb - dataset parsing
• models - answering models
• • elasticsearch_baseline - fuzzy search
• • bert_emb_baseline - BERT embedding ranking model
• • bpe_baseline - BPE embedding ranking model
• • use_baseline - USE embedding rankinig model
• Run bot python chatbot.py
• Available in telegram by address @ods_help_bot
• Available in Slack by address digestbot QnA functionality

Features

• answer question
• show other relevant answer (top 4)
• estimate question (Good, Bad) only in telegram by now
• show history

Running

• You need your own telegram TOKEN from BotFather
• You can run bot in screen
• add token to system variables export TOKEN="somerandomtoken"
• create virtual environment python -m venv venv (optionally)
• pip install -r requirements.txt
• installed elasticsearch
• running systemctl restart elasticsearch
• python chatbot.py

Finite automaton