Predict whether a news article will have a significant impact on the underlying crypto-asset mentioned in the article

Contact: Anand Gomes, [email protected] & Elias Humberto, [email protected]

Introduced by the team at Paradigm. About Paradigm: We are powering the $230+ Billion OTC crypto market by building a conversational interface for institutional traders. Our mission is to increase revenue and efficiency which makes traders lives easier by providing AI-driven tools such as automated trading and counterparty discovery within a native chat application.

Problem: The crypto-news cycle generates an overwhelming amount of news, both fake and true. There is no reliable and inexpensive method of filtering which articles will significantly affect the price of underlying crypto-asset it refers to. Traders need an automated tool that is able to read news articles real-time and dynamically assign a score based on the probability of impact on price.

Bloomberg and Reuters, two financial data/news powerhouses cost up to $25K/year for their news feed but do not solve this problem and instead only provide you simple metrics such as whether an article is trending or if it has clocked the most reads over the last day/month/week.

Solution: A beta version of the project already exists. Teams will build on the work done by the Berkeley team from last semester under the guidance of a project lead (Elias Castro – ML Lead), industry mentor (Anand Gomes, CEO of Paradigm), with input from a senior researcher at Uber’s AI Labs. Note that this project will focus on the impact of the story on price and NOT on whether the news is real / fake.

Datasets

1. NewsAPI: A JSON API for live news and blog headlines that aggregates news articles from multiple news sources

2. CoinAPI: Subscription to CoinAPI’s cryptocurrency market-data