Inspirata is a crowdsourced platform, which extracts motivating and encouraging tweets from Twitter and uses them to train a Markov Chain to produce encouraging words for those who might show signs of depression, suicide, anxiety etc.

Updates We used NLTK's part-of-speech tagger to generate a Markov model that obeys sentence structure better than a naive model.
Responses generated by Inspirata can be seen here and also on Twitter and program is running on a local server.

As an update to our platform, we extended our idea to instagram, where we made a service which searches for certain keywords on instagram and then using sentiment analysis on captions and emotion analysis on images, we find if the instagram post is sad or depressing. Then we comment on that post with encouraging words.

We used Microsoft Text Analytics APIs for Natural Language Processing and then used a markov chain to produce sentences. We first use Language Detection to segregate tweets in English. Sentiment analysis then calculates positive or negative sentiment which is used to classify the tweet as sad or happy. If the tweet is a sad one, our app replies the person who tweeted with an inspiring qoute. If the tweet is a happy one, our app uses it to train the Markov chain model for generating inspiring quotes to tweet. In this sense, our app is a crowdsourced app. For Instagram we used the same algorithm but added additional feature to use an image for Emotion Detection that further enhances its functionality.

Above two pictures are sentences from our file on which markov chain is trained, and as we can see how the below sentence is generated.

Currently inspirata updates a tweet with a 20-30 minute interval.