NLP Techniques for Financial Investment Analysis: Case Studies from Two Sigma, BlackRock, UC Berkeley and Others

Financial analysis seeks excess returns by exploiting information asymmetry, which increasingly comes from massive unstructured data such as news, social media and research reports. Natural language processing (NLP) enables investors to extract valuable signals from these texts quickly and efficiently.

NLP in finance originated in the 1980s, with Google’s 2003 patent demonstrating the predictive power of news. The rise of social media in 2011 highlighted the importance of public sentiment, and deep‑learning models such as CNNs, LSTMs and word embeddings have further enhanced text understanding.

Typical NLP tasks for finance include:

Keyword or topic extraction using bag‑of‑words, LDA or modern embedding techniques.

Sentiment analysis to gauge market mood.

Event extraction via templates, syntactic parsing and named‑entity recognition to identify corporate actions.

Case Study 1 – Topic Modeling (Two Sigma): Two Sigma applied LDA to decades of FOMC minutes, identifying eight topics and tracking their evolution. The analysis revealed shifts such as reduced discussion of growth and increased focus on financial markets, providing investors with macro‑economic insights.

Case Study 2 – Event Extraction (UC Berkeley & BlackRock): Researchers extracted events from news headlines using subject‑verb‑object patterns, mapped companies to the S&P 500, and clustered similar events with k‑means. Event‑driven portfolio tests on 2006‑2014 training data and 2015‑2016 out‑of‑sample data demonstrated significant excess returns, especially for “oversold conditions” and “approval” events.

Case Study 3 – Sentiment Analysis (Twitter, 2011): Using OpinionFinder and GPOMS, the study measured seven emotional dimensions from tweets and applied Granger causality and a self‑organizing fuzzy neural network (SOFNN) to predict the Dow Jones Industrial Average. The “Calm” dimension showed strong predictive power, while overall polarity did not.

The review concludes that while NLP has become a crucial tool for financial analysis, many applications still lag behind the latest research by several years, largely due to a shortage of professionals who understand both finance and advanced NLP. Deeper collaboration between domain experts and AI researchers is essential to unlock the full potential of modern NLP in investment strategies.