With the 2016 US Presidential election just around the corner, our news feeds are overflowing with information and conflicting statements, each side claims are facts. How can anyone know what to believe? We wanted to build a tool that could bring clarity to any conversation by checking facts in real time.

What it does

Imagine you're watching the presidential debate, and you hear one of the candidates throw out a dubious statistic or make a outrageous accusation against their opponent. Prompt FactCrunch to check the statement's validity by saying "Ask factchecker, is that true?" FactCrunch investigates the last statement it heard and responds with a verdict that let's you know how true or false the statement is. FactCrunch does the research for you, so no need to scour the web for answers.

How we built it

We used IBM's Watson to convert speech to text to process statements being made in real time. Using Amazon Alexa, we created a trigger for users to initiate a search on the most recent thing they heard. Alexa then communicates with an API service we built with nodeJS and express, that queries trusted and vetted news articles from Politifact with a custom google searrch to determine a ruling for the statement. Alexa's voice then instantly responds with a ruling and summary of the statement.

Our website will live at which we've purchased.

Challenges we ran into

It was very cold at Pier 48! All of the API's used were new to our team. Our data sources were difficult to isolate, and we ran into some rate limiting with our custom searches.

Accomplishments that we're proud of

We were able to listen to an audio stream in real time, and capture the relevant statements when a user triggerd a fact check. The interaction experience is very natural and we believe that this type of tool could be a great companion for people interested in the truth, the whole truth and nothing but the truth.

What we learned

As we built and tested our app, we definitely were set straight on a lot of facts! The earth is definitely not flat for one! We learned a ton about the IBM Bluemix platform and how it offers amazing enterprise tools to any developer. We learned how to create Alexa intents and utterances, and how to connect Alexa's and IBM Watson's AI's to broader data sets and uses.

What's next for FactCrunch

We'd love to add more fact checking sources, and also give users the option to follow up and learn more about each statement if they're interested in digging deeper. We'd also love to have FactCrunch integrate with video streams and display relevant information inside a video, and explore ways to gameify the app (maybe display a Pinocchio meter under each candidate for example), or build a trustworthiness indicator for whenever a speaker is in question. Also we'd like to build a homepage for the app.

Share this project: