Studies, including an experiment by Facebook, have proven that the content of your news feed can affect your emotions. An overly-negative feed has the potential to cause depression.

We tackle this problem by scraping users' Facebook feeds and assigning them a negative-positive score using HP's Haven OnDemand Sentiment Analysis API. If this score is too low, we alert the user that their feed may be aversely affecting their mood. We can then replace negative posts with cute gifs (guaranteed to improve anyone's day!) or alert a loved one/therapist about the users' exposure to potentially triggering content. A (subtle) link is provided to reveal the original content.

Our front-end chrome extension parses each post on the user's Facebook newsfeed, concatenates it with any comments, and passes the text to HP's Sentiment Analysis API to get an aggregate "positivity" score. This positivity score is then logged in a MySQL database in one-hour increments, and any trigger-words are identified. The decision to filter a post is calculated based on the last 12 one-hour blocks and the frequency of certain trigger words.

We experimented with HP's Entity Extraction API to look for specific types of content in the posts, for instance searching for content associated with particular themes that could be a trigger for users, but in the end we didn't have time to integrate the API. Instead, we check for particular triggering themes by searching through the given post text for groups of related words associated with these triggers. This ended up serving us well, giving us a solid categorization of text into different themes of some of the most common triggers in people with past traumatic experiences.

Share this project: