Computer-Aided Patent Analysis: finding invention peculiarities

The application of standard Information Extraction techniques to Patent Analysis has several limitations partially due to the difference existing between patents and web pages, which are the object of the biggest majority of information search. Indeed, while in other fields customized processing techniques have been developed, the number of studies fully dedicated to patent text mining is very limited and the tools available on the market still require a relevant human workload. This paper presents an algorithm to identify the peculiarities of an invention through an automatic functional analysis of the patent text; as a result a ranked list of components and functions is provided as well as a selection of meaningful paragraphs disclosing the details of the invention. An example related to laser irradiation devices for medical treatment clarifies its basic steps.