In discovery, document review is one of the most mundane tasks. With this application, one aspect of document review can be accelerated by helping the user identify texts that have a high emotional score.
How it works
Our emotion score algorithm is based on: 1) swearing, 2) punctuation deviation, 3) smilies, 4) short sentences, 5) long sentences, and 6) keywords. These indicators are tallied up and divided by the lines or number of sentences to reach the emotion score.
Challenges I ran into
Please be sure to use Python 2.7.10 to run this application. We had challenges identifying sarcasm and sexual undertones.
What's next for Emotion Detector
To improve accuracy, we would need to collect data to be able to help the application understand context in which words are used and how frequently they are associated with a certain emotion.