Semantic exploitation of implicit patent information

Abstract

In recent years patents have become increasingly important for businesses to protect their intellectual capital and as a valuable source of information. Patent information is, however, not employed to its full potential and the interpretation of structured and unstructured patent information in large volumes remains a challenge. We address this by proposing an integrated interdisciplinary approach that uses natural language processing and machine learning techniques to formalize multilingual patent information in an ontology. The ontology further contains patent and domain specific knowledge, which allows for aligning patents with technological fields of interest and other business-related artifacts. Our empirical evaluation shows that for categorizing patents according to well-known technological fields of interest, the approach achieves high accuracy with selected feature sets compared to related work focussing on monolingual patents. We further show that combining OWL RL reasoning with SPARQL querying over the patent knowledge base allows for answering complex business queries and illustrate this with real-world use cases from the automotive domain.