Tweet Reply Sentiment Application

This is an application that can scrap twitter replies and predict a sentiment. It also provides various graph visuals.

Installation:

You can use the application by going to https://tweet-sentiment-application.herokuapp.com/. If you'd like to use it locally then download the following files: app.py, model.pkl, prediction.py, text_cleaner.py, tweet.py, user.py, visualizer.py, Optional: amazon_test_data.csv.

Using a terminal type:

 streamlit run app.py

File descriptions:

Data = contains the main data used to train the model 

Processed = contains the output data of the data_processing.py script

Procfile = used by heroku to run the application

amazon_test_data.csv = around 1000 tweets scraped from the amazon twitter account. Use this to test the application if you don't have API credentials

app.py = the main streamlit app

data_processing.py = the script used to clean and prepare the data to train the model

model.pkl = the trained model

model.py = the script used to train and test the model

nltk.txt = for Heroku to access stopwords

prediction.py = used in model.py, a class that contains the different sklearn models and their settings

requirements.txt = the required packages

text_cleaner.py = the class used in app.py to clean tweets

tweet.py = the class used in app.py to access twitter API and retrieve replies

tweet_info.jpg = a visual to show how to find tweet information

user.py = the class used in app.py to authorize twitter API access 

visualizer.py = the class used in app.py to provide data visuals, bar graphs and key word searching.