isc-hemc / burrobot

Burrobot is a prototype of conversational agent popularly known as chatbot that will interact with users using Artificial Intelligence and NLP techniques integrated into Facebook Messenger network.

Burrobot

Burrobot is a prototype of conversational agent popularly known as chatbot, that will interact with users using Artificial Intelligence and NLP techniques integrated into Facebook Messenger network.

Contents

• Dependencies
• Structure
• Configuration
• Run
• Authors

Dependencies

Structure

Before start diving in, the project it's defined in different modules:

• qa: stores a Flask API that turns out to be the Burrobot service itself. This API has a pretrained model using transformers and BERT.
• cms: stores a Django API that will be used to update prone to changes information such as dates or any other business data.
• docs: stores everything that has to with the thesis, the investigation behind it and the results.
• mongo: holds the MongoDB configuration that Docker needs to run it properly, customly and safely. In here we will store the raw messages comming from the Facebook API.
• mysql: holds the MySQL configuration that Docker needs to run it properly, customly and safely. In here we will store the normalized information of the messages that we previously store in mongo with the purpose of use it to calculate information such as most important topics and plot them to evaluated which ones the community are talking about.
• nlp: this module is dedicated to Natural Language Processing tasks. In here we implement the normalization, topics analysis, LDA and plot tasks.
• scraper: as its name says, holds the implementation of a scraper to download facebook messages from a profile thanks to Facebook's Graph API.

Configuration

Before running any module (other than documentation) it's recommended to build and run the MongoDB and MySQL databases with the help of docker-compose, for the other modules inside its folder is a README file that gives an extended explanation of the purpose of the module and how to configure and run it.

To build properly the docker images it's necessary an environment file for each database.

MongoDB

Inside the mongo module, open a terminal and type the following command:

touch mongodb.env

Once the mongodb.env file it's created add the following environment variables:

Variable	Value	Description
MONGO_INITDB_ROOT_USERNAME	root	Creates a new user in the admin authentication database and gives the role of root
MONGO_INITDB_ROOT_PASSWORD	pass	MONGO_INITDB_ROOT_USERNAME password
MONGO_INITDB_DATABASE	mydb	Specify the name of a database to be used for scripts in /docker-entrypoint-initdb.d/*.js

MySQL

Inside the mysql module, open a terminal and type the following command:

touch mysqldb.env

Once the mysqldb.env file it's created add the following environment variables:

Variable	Value	Description
MYSQL_ROOT_PASSWORD	rootpass	Specifies a password that is set for the MySQL root account
MYSQL_USER	user	Creates a new user and grants superuser permissions for the database specified by MYSQL_DATABASE
MYSQL_PASSWORD	pass	MYSQL_USER password
MYSQL_DATABASE	mydb	Database to be created on image startup

Run

Once the environment files for each database are in place, open a terminal and type the following commands:

Build the images specified in docker-compose.yml

docker-compose build

Creates and run the Docker containers for each image

docker-compose up -d

With these, both database should be running healthy.

Authors

*David Hernández Martínez - Davestring

*José Ricardo López García - JoseRicardoL

*Iván Giovanny Mosso García - igmosso84

*Pabel Carrillo Mendoza - pab3l