Twitter Sentiment Analysis Bot

Summary

A simple Twitter bot written in Python that will perform sentiment analysis of tweets from specific user by request.

Installation

The bot can be run locally or deployed to the hosting. This repository contains all required files for the bot to be deployed to Heroku. In order to run successfully, four environment variables must be set on a target machine:

• twitter_sent_bot_key that contains Twitter API key

• twitter_sent_bot_secret that contains Twitter API secret key

• twitter_sent_bot_token that contains Twitter access token

• twitter_sent_bot_token_secret that contains access token secret

Requirements

The bot is written in Python and requires the following dependencies:

• tweepy == 3.6.0

• numpy == 1.15.1

• pandas == 0.23.4

• matplotlib == 2.2.3

• vaderSentiment == 2.5

Description

The bot is launched by using console command python App.py. Additionaly, an integer argument can be provided that defines the total running time in seconds (default is 360 seconds = 5 minutes). The bot checks every minute for all mentions of the current user and if it detects a tweet where it is mentioned along with another user (target user), it then loads up to 500 latest tweets of the target user, performs a sentiment analysis of them using VADER, plots a line chart of the compound score and replies back to the original tweet, posting this chart. The bot tries to prevent abuse by scanning each account only once. Below are some example of these charts

• Tweets of @HellthyJunkFood

• Tweets of @k_huck

• Tweets of @UCIrvine

Notes

Keep in mind that Twitter will most likely not return mentions that are done by users who are likely not human e.g. if a user posted something via Twitter API, did repetitive tweets or is just a new user with next to no tweets of follows.

P.S. The bot was developed and tested under @ChaplyginAndrei account, follow this link to see more examples