Cinder

Inspiration

We set out to create a smart curation system that could influence the kind of news briefers receive each day through their interactions with the previous day's briefs.

What it does

Our product is a briefing dashboard for intel curators and briefers. It captures the substantial knowledge of intelligence professionals with an approach that is both scalable and systematic.

It uses four points to track user interest:

Did they put this in their brief? Did they read this article? Did they ignore this? Did they dismiss this?

Briefs added to the book are rated as "most relevant," while articles read are marked as "more relevant." Articles passed over are rated as "less relevant," and articles dismissed are "least relevant."

In this way, different users will get different articles presented to them based on their interests and previous interactions.

Finished books can be printed out or presented to customers in a digital tablet readout. If presented digitally, our product will one day capture the time a customer spends on each article, and results are once again rated by our algorithm.

How we built it

The brains of the operation is a custom implementation of the Naive-Bayes model. Topics of interest are automatically generated using TFIDF vectorization of article text, eliminating the need for manually-generated tags. The ultimate goal is to alleviate the workload of the curatorial staff, by allowing briefers to quickly and effectively create books for their customers based on continually refined results.

Built With

• express.js
• naive-bayes
• node.js
• python