Today's political climate is full of energy and strong opinions. But it's often difficult to channel this energy in an meaningful way. The knowledge bias is also very prevalent due to echo chambers throughout social media. So, we wanted to build a product that properly informs citizens about different perspectives in our political landscape, encourages them to reflect on their values, and get more involved in the political process.
What it does
Doug provides a political news digest to the user upon request, and can even take search queries to further provide content related to a particular topic. It can provide bite-sized summaries of the latest events and policy decisions. Based on the citizen's interests and values, they can seamlessly contact their Senators, donate to causes, or get involved in local activism to make a difference. The end goal is to have Doug be a reliable friend that one can vent to and receive insight from.
How we built it
We used a Python/Django server that connected to an api.ai NLP engine, Facebook Messenger interface, machine learning models, and Microsoft Cognitive Services.
Challenges we ran into
Most of our challenges arose from the integration of many different pieces of software. Furthermore, there are always unlimited edge cases in conversations with human beings. We strived to address as many of them as possible.
Accomplishments that we're proud of
The big accomplishment is in the complex coordination of many different systems in order to make the experience so seamless. This product relies on over five third-party services to deliver accurate information and increase awareness on action items. Most importantly, we're proud of the impact that such a product can have on millions of people who are either really optimistic or really concerned about the future of the country.
What we learned
We learned how to train our internally developed classification engine to categorize political articles as related to Immigration, Economy, Climate, Education, or Healthcare. Additionally, we learned how to use the Facebook Chatbot SDK and connect that to an NLP system to discern intents and actions of the user.
What's next for Doug
We'd like to further train the NLP engine to have Doug become even more conversational in the future. Furthermore, there is lots of scope to increase Doug's capabilities on the awareness and action front. We would like to have Doug provide spontaneous factual answers to follow up questions posed by users, and also generate ad-hoc answers & analyses upon request.