UBC-NLP / EmoNet

EmoNet

Introduction

EmoNet is a neural network tool for emotion recognition, described in this paper. The current packaged version predicts 9 emotion categories (p2 in the paper). This is a work in progress, and so the rest of categories will be added. ... [TBD]

Installation

This repo is tested on Python 3.6+, PyTorch 1.2.0+ and TensorFlow 2.0.0

First you need to install both, TensorFlow 2.0 and PyTorch .

With pip

[TBD]

From Source

Once TensorFlow 2.0 and PyTorch has been installed, you can install from source by running:

pip install git+https://github.com/UBC-NLP/EmoNet.git

It will automatically install other dependencies including ' happiestfuntokenizing', 'transofrmers', 'numpy' and 'pandas'.

Usage

As a Python Library

You can use emonet as a Python library.

EmoNet.predict

Parameters	Description
text	Specifies the string for prediction
path	Specifies the file path to the tsv file for prediction. (It must have a 'content' column)
with_dist	Specifies whether diplay the distrubution over all class labels

Example:

from emonet import EmoNet

em = EmoNet()

# predict a text language and returning the distribution over all languages
prediction = em.predict(text='Spectacular day in Brisbane today. Perfect for sitting in the sun and thinking up big ideas and resetting plans.', with_dist=True)
print("Predict a text and display the distribution:")
print(prediction)

# Predict text in a tsv file line by line
predictions = em.predict(path=path_to_tsv_file)
print("Predict a text file line by line:")
print(predictions)

Here is the output:

Predict a text and display the distribution
[('joy', 0.8978073, {'anger': 0.0008576517, 'anticipation': 0.06090205, 'disgust': 0.00068270933, 'fear': 0.007252514, 'joy': 0.8978073, 'sadness': 0.004249889, 'surprise': 0.025819499, 'trust': 0.0024283403})]

Predict a text file line by line:
[('joy', 0.9871133), ('anger', 0.94085765), ('fear', 0.99755955), ('anticipation', 0.98000574), ('joy', 0.5602796), ('joy', 0.35310036)]

As A Command-Line Tool

emonet.py [options]

Options:

-b, --batch	specify a file path on the command line
-d, --dist	show full distribution over languages

Example:

Assuming path/to/file/text.tsv containts the following:

content
I am happy about this 
That movie was scary

then running:

$ emonet -b path/to/file/text.tsv --dist

gets you the output:

[('joy', 0.85855854, {'anger': 0.009707213, 'anticipation': 0.011866086, 'disgust': 0.00912305, 'fear': 0.018755471, 'joy': 0.85855854, 'sadness': 0.039343264, 'surprise': 0.037687544, 'trust': 0.014958782}), 
('fear', 0.6364692, {'anger': 0.02605511, 'anticipation': 0.020012422, 'disgust': 0.013645536, 'fear': 0.6364692, 'joy': 0.09334463, 'sadness': 0.14103319, 'surprise': 0.06486984, 'trust': 0.00457005})]

Interactive Mode

It is very simple to use the interactive mode. Invoke using python emonet.py or just emonet.py if you have already installed the package.

~> python emonet.py
>>> Spectacular day in Brisbane today. Perfect for sitting in the sun and thinking up big ideas and resetting plans.
[('joy', 0.8978073)]
>>>

You can also specify a file path by using --batch option. The script also detect when the input is redirected.

python emonet.py < test.tsv
[('joy', 0.8978073)]