Technical problem identification for supervised state of the art

This paper presents a method for extracting technical information from a patent pool. It was designed to support the construction of the state of the art of a technology or a product/process by automatically identifying the list of problems that the inventors have faced. The method is based on a strict ontology, which defines what a patent problem is, and a set of IR strategies, which identify all alternative ways adopted in the pool to describe problems. More in detail, the authors propose a set of syntactic dependency patterns, and lemmas in order to extract only the sentences including the information dealing with problems. The output is a coarse list of technical problems, automatically extracted without the user being an expert in the problems of the sector. An exemplary case dealing with injection molding field is proposed.