Sep. 14 2017 — Unveiling The Hidden Information In Earnings Calls
Given the growing interest in NLP among investors, we are publishing this primer to demystify many aspects of NLP and provide three illustrations, with accompanying Python code, of how NLP can be used to quantify the sentiment of earnings calls. In our first example, sector-level sentiment trends are generated providing insights around inflection points and accelerations. The other two illustrations are: i) stock-level sentiment changes and forward returns, and ii) language complexity of earnings calls.
- What is NLP? – We demystify common NLP terms and provide an overview of general steps in NLP.
- Why is NLP important? – Forty zettabytes (10^21 bytes) of data are projected to be on the internet by 2020, out of which more than eighty percent of the data are unstructured in nature, requiring NLP to process and understand.
- How can NLP help me? – We derive insights from earnings call transcripts via NLP measuring industry-level sentiment trends or language complexity of earnings calls, and much more.
- Where do I start? – Code for each use case is enclosed, enabling users to replicate the sentiment analysis.
Natural Language Processing, Part I: Primer
Download the full report
David Pope, CFA, S&P Global Market Intelligence’s Managing Director of Quantamental Research, recently discussed using natural language processing to unlock new insights in corporate earnings sentiment analysis. Click the player to view the video.
Natural Language Processing – Part III: Feature Engineering