High-content Words in Patent Records Reflect Key Innovations in the Evolution of Technology

We study the evolution of technology as reflected in the US utility patents granted in the period 1976-2009. Previous work by Skusa and Bedau (2002) and Buchanan et al. (2010) used cumulative citation statistics to identify the inventions that most affect the course of evolution (those with the highest innovative impact). Here we examine the text of patent records (specifically, titles and abstracts) to identify which features are responsible for the high impact on later innovations. We use the TFIDF metric (term frequency times inverse document frequency) to identify which words best convey a patents explicit content. Because a new patent is required to cite all important earlier patents (“prior art”) that introduced innovations on which the new patent depends, we use the TFIDF scores of words in citing patents to identify a patent’s emergent content. A patent’s emergent content explains its impact on subsequent inventions; it reflects what traits in an invention actually led to a significant number of subsequent innovations. We illustrate two ways to visualize the explicit and emergent content of patents: word arrays and clouds. Examining the emergent content of populations of patents issued during different epochs reveals when important new ideas appear in the evolution of technology and how they affect its subsequent evolution.