Neural Network for Natural Language Processing: Emotion classification on Twitter

A machine-learning approach to detect emotions out of tweets.

You can send an email at: [email protected] for the different dictionaries, or the original tweet corpus if you want to build your own.

Python dependencies:

• numpy
• tensorflow

Usage

You can find the pre-trained embedded vectors on the stanford.edu platform: https://nlp.stanford.edu/projects/glove/

It is recommended to download the four pre-trained vector set trained from Twitter (dimensions 25, 50, 100, 200).

---

All the tensorboards results are stored in the /runs dir.

model.py describes the convolutional neural network, and classifier.py is the glove implementation.

Flags:

• -b: embeds the training_data and the test_data. It saves it in the dir /bin, so you might want to create it first.