You can find the deployed application at : https://tweet-emotion-detection.herokuapp.com/
Code for the heroku app : https://github.com/karthikchiru12/Tweet-emotion-detection-heroku-app
About :
This Project and the report are built as a part of a hackathon, conducted by Spotle.ai. As a part of this project, we are provided with a set of tweets. And the task was to find a way to associate or classify a tweet with a particular emotion.
We have done two experiments, in one we used hashtags and emojis to label a tweet with its emotion. And in the next experiment, we tried to label the tweets based on a lexicon database of words with emotions. And after the project, we concluded that the first way of labeling was not successful. But in the second approach, we were able to successfully label the tweets with the emotion using contextual emotion vectors of words in lexicon database.
We have not used any external dataset of tweets in this project.
The lexicon database used in this project is downloaded from the NRC (National Research Council Canada) and the lexicons used in this project are created by the National Research Council Canada.
http://sentiment.nrc.ca/lexicons-for-research/NRC-Emotion-Lexicon.zip
Future plans :
- Transform the code from jupyter into a package
- Train on a bigger dataset