Multi-dimensional reputation system for Twitter

Name of Project:
Replabs: Multi-dimensional reputation systems.

Proposal in one sentence:
Building a multi-dimensional reputation system for determining legitimacy on social media using NLP.

Description of the project and what problem is it solving:
With the internet, trust in universities and newspapers is fading while the amount of content we’re interacting with is exploding. This void of legitimacy is filled by reputation systems that collapse reputation into likes, retweets, stars, karma-points, tokens or hyperlink density. Combined with engagement-optimizing algorithms, credibility has become a function of virality. As a result, we are struggling to know who to trust, who to follow, and how to make sense of the world.

Replabs aims to solve this by building multi-dimensional reputation systems using language models.

For this month, the focus is to create a fully functional POC chrome plugin for Twitter. The plugin will scrape Twitter data continuously, and create reputation graphs from this using PageRank and language embeddings. This information will then be overlayed on twitter profiles, allowing users to more easily see who is trustworthy in controversial conversations.

More information about Replabs and our mission can be found here:

Grant Deliverables:

Grant Deliverable 1 – Twitter chrome plugin POC
Grant Deliverable 2 – Technical paper draft (will be finalized at