Sentiment Analysis of Tweets for depression

Inspiration

Around 800,000 people suffer from depression every year and around half of them are left uncured, thus increasing the sucide rate. Sucide rates are 2nd highest cause of death for people of 15-29 years age. With this project we are trying to help the people that are left uncured or undetected.

What it does

It uses NLP to predict if a tweet is depressing or not. It takes 8 latest tweets and then take average of the prediction percentage and then check the average to see if the person is depressed or not.

How we built it

We used CNN and LSTM models to make this.

Challenges we ran into

• It was very difficult to collect data for this problem.
• The complexity for LSTM model was very high and thus it took a long time to train this model. ## What's next for Sentiment Analysis of Tweets for depression We would like to add a twitter bot that notifies the person and their loved ones to get help. We would also like to apply this to other texting platforms like Whatsapp, Texts etc.

Built With

• keras
• python
• tensorflow