Hands-On Machine Learning for Algorithmic Trading

RavenPack

RavenPack analyzes a large number of diverse, unstructured, text-based data to produce structured indicators, including sentiment scores, that aim to contain information relevant to investors. The underlying data sources range from premium newswires and regulatory information to press releases and over 19,000 web publications. J.P. Morgan tested a long-short sovereign bond and equity strategies based on sentiment scores and achieved positive results with low correlation to conventional risk premiums (see references).