bhyvex / fib-chatbot

A chatbot developed to help FIB students. It uses both NLP and Machine Learning techniques to answer questions.

Fibot

Fibot is a telegram bot that is able to help FIB students through conversations. You can try him online by going here!

Libraries

It is written using:

• python-telegram-bot: https://github.com/python-telegram-bot/python-telegram-bot
• rasa_nlu: https://github.com/RasaHQ/rasa_nlu
• rasa_core: https://github.com/RasaHQ/rasa_core

Table of contents

• Quick start
• Setup

Quick start

After getting everything set up (more about how in Setup ...) it is only necessary to run the run.py script.

It will automatically boot up the bot and it will start to be effective and get to respond queries from students.

Setup

Make sure to have everything installed. Notice it is necessary to have the corpus for both 'en' and 'es' languages on spaCy. You can download them by using the following commands on your terminal.

  python -m spacy download en

and

  python -m spacy download es

Then, there is a step necessary to execute everything correctly, which is to set up several environment variables:

• client_id: with the client_id value obtained by registering your app in here.
• client_secret: with the client_secret value obtained by registering your app in here.
• encryption_key: a 16-character-long number that acts as private key for encrypting sensible data.
• FibotTOKEN: The Telegram bot Token as obtained by Telegram's BotFather here after creating the bot.

Next, if you have not downloaded the models folder with the pretrained files, it is mandatory to execute first the train_models.py, so that it can train the necessary models and store them to be used later.

Also, it is possible to automatically generate the dataset that the rasa_nlu is trained on using the generate_dataset.py python script, that allows to create a dataset for any language of the three of a fixed size randomly (using the data in Data/Professors.txt and Data/Subjects.txt, etc).