Analyzing Automation Technologies and Their Tasks in Patent Texts Using Natural Language Processing

Technological advances have changed the labor market, and automation technology is a major factor affecting jobs. To study its impact, it's important to accurately measure automation technologies. Patent text is a novel approach to understanding technological progress and identifying the tasks that different automation technologies can perform. The paper uses a dictionary-based approach to identify automation technologies from patent text and categorizes them into three groups: Robots, Software, and Artificial Intelligence. We find that the number of patents related to these groups has grown exponentially since 1980, with AI patents growing the fastest since the 1990s. Most patents are in the Physics (G) patent family. The tasks performed by the different groups of automation technologies vary, with robots focusing on assembly tasks, software on information processing, and AI on higher cognitive tasks. Almost all technologies aim to improve efficiency and reduce costs, according to the patent description.