Tweet Sentiment

About

• A small app which uses a pre-calculated model to classify tweets.
• The notebook tweet-sentiment.ipynb shows how I trained the model.
• The model tweet-sentiment.pkl classifies text into a category: negative (0), neutral (2), positive (4).
• The purpose of this project was to gain some experience using fast.ai - text classification.
• I am using the sentiment140 data set.
• For more information see fast.ai lessons: 4 here, this post also helped me to create the classifier.

Requirements

• Docker

Installation

Api

// build and run app
./gradlew build run

// stop app
./gradlew stop

Twitter

Add your CONSUMER_KEY, CONSUMER_SECRET, ACCESS_TOKEN and ACCESS_TOKEN_SECRET into twitter.env.

Notebook

You will need access to a GPU to run the jupyter notebook. Fast.ai recommend using a p2.xlarge instance. Follow the installation guide here.

Usage

User interface

localhost:8000