Facebook Emotions

Download your messages

Choose a friend

And reveal all emotions end-to-end

Setup

Download the zip file of your data from here.

You need to tick the message checkbox.

Then you need to set the format to JSON and create a file.

It may take a few hours for Facebook to process your request.

Once your zip file has downloaded:

• Clone the repository
• Install the requirements
• Unzip your data into the repository

Usage

Fake Data

Before running the code on your own data, you can check everything is working on the fake data.

>>> python src/main.py -p fake_data -f johnsmith

This should create the following image in FacebookEmotions/word_clouds.png

Your Data

If this ran successfully, you can run the code on your own data.

>>> python src/main.py -p PATH -f FRIEND -l LIMIT

where

• PATH is the path to your unzipped Facebook data.
• FRIEND is the full name of a friend (I.e. Johnsmith).
• LIMIT is the maximum number of message to process. On my machine, 1000 messages can be processed in about a minute.

Explanation

This project is based on the Ekman model found in https://github.com/monologg/GoEmotions-pytorch.

It can be found in Hugging Face under "monologg/bert-base-cased-goemotions-ekman".

I could not find any other good models that included the 'neutral' emotion, so please get in touch if you find something more appropriate!

Possible improvements

• Create our own transformer model with greater accuracy.
• Distill the model so that it can perform inference faster.