OpenMind

Inspiration

We got annoyed with the way today's recommender systems work - digging a rabbit hole for the user consuming content which results in the user seeing homogeneous recommendations with minimum variety just to drive more engagement. When it comes to highly-opinionated topics, this often becomes dangerous for the users strengthening their opinion on a topic without being challenged.

What it does

The chrome extension recognises when user reads a highly-opinionated content, conducts NLP sentiment analysis of the content and offers a content recommendation with a opposing view/challenge via pop-up.

How I built it

TLDR we duct taped a bunch of APIs together and hoped for the best.

Our chrome extension detects text on screen and serves it to GCP Cloud Function which forwards it to Google natural language API which does the heavy lifting of analysing article contents. The Cloud Function then uses the most salient keywords identified by the NLP engine to search for articles on the same topic using the webhose.io API. The search results are again analysed by the NLP engine. The Cloud Function then identifies a related article with the most differing opinions that gets displayed in the Chrome extension pop-up.

Challenges I ran into

Evaluating recommended articles

Accomplishments that I'm proud of

Creating end-to-end ready to use product (OpenMind chrome extension)

What I learned

Not to push private API keys to a public GitHub repo

What's next for OpenMind

funding and one-way flight to silicon valley, naturally

Built With

• flask
• google
• javascript
• python
• webhose