An Analysis of the Patent Search Behavior of Human Experts and its Impact on Patent Retrieval Performance

Patent databases are considered valuable sources of technical information, yet, retrieving patents from databases is viewed as a tedious and inaccurate process. Currently, patent retrieval is primarily conducted by human experts, whose performance at this endeavor can vary greatly. Different approaches to how human experts search for patents in patent data bases may constitute the source of this variability in performance.This paper describes a study which investigates the impact of patent search behavior on patent retrieval performance. A researcher observes seven experts in patent retrieval as they search for, extract and judge metallurgical patents that could contain enabling technologies for developing and producing a new and improved kitchen skillet. The researcher measures search behavior in terms of keyword diversity, query complexity and search speed. He evaluates retrieval performance against well-known measures for reliability, efficiency, effectiveness and judgement error. The study clearly establishes a relationship between patent search behavior and patent retrieval performance. It also demonstrates that the patent retrieval performance of human experts is very low and varies greatly. The results of the study suggest that applying artificial intelligence methods to patent retrieval could improve patent retrieval performance immensely.