In this project, we use dynamic word embeddings and other NLP techniques to trace the change of business environment for U.S. companies in the past 45 years. We use over 100 business publications and patent corpus to build a model for the dynamic business landscape, which contains rich information about potential commercial opportunities and relational business concepts. The outcome is a multi-slice embedding space, with each slice reflects the underlying economic reality and Zeitgeist of the time.

By placing industries and individual organizations in this context, we build rich measures to evaluate their combinatorial novelty and potentials for future success. Our method helps us break the assumption of a static, ahistorical market, and help locate readers in the historical contexts where an entrepreneur makes her business decisions. Through the computational lens, we trace the complexities in business evolution.