Unihack 21 Submission -- Team 14

Refactum

Inspiration

In the past year, we observed disinformation surging. There’re spates of new stories and news every day, but in addition to the authenticated reports, many rumors are flying around. We ask ourselves — how do we identify false information in unscrupulous media more efficiently at scale? We started from Allen, Jennifer N. L., et al.’s paper “Scaling up Fact-checking Using the Wisdom of Crowds.” (PsyArXiv, 1 Oct. 2020. Web.) and introduced the idea of knowledge graph into fact-checking.

What it does

We emphasize the rigorous presentation of evidence and the tree of evidence, to ensure, using algorithms and HCI design, both the efficiency and accuracy of fact-checking. Users are invited to share, connect, and argue about facts/myths.

How we built it

Front-end:
- Typescript + React + Chakra UI + Grpc-Web
Back-end:
- Go + Protobuf & Grpc + BadgerDB + Grpc-Web
DevOps:
- Docker + Kubernetes + Envoy Proxy
PM:
- Agile

What's next for Refactum

Scalability
- Since our data are stored in key-value pairs, we can use blockchain technology to make our store distributed across multiple immutable stores
- Developed with Go & Grpc → prepared for microservice and large scale orchestration
Prospects
- Natural language processing to learning document embeddings
- Graph Neural Networks (GNNs) to deep-learn the hidden representations of fake content
- Partnership with big companies to call for public engagement