Automatic Categorization of Patent Documents for R& D Knowledge Self-organization

The World Intellectual Property Organization (WIPO) reports that ninety to ninety-five percent of all R&D refers to existing patent documents. A company can reduce costs and shorten development time by effectively utilizing existing knowledge, as disclosed in the global patent corpus and in the intellectual property news media. As a consequence, patent information plays an important role in the era of knowledge-based economies. However, owing to the dramatic increase in the number of patent documents people have difficulty reading, organizing, and fully utilizing them. There are also unique technical and legal vocabularies in the context of patent documents that prevent adequate understanding of patent claims. The consistent and effective organization of important ontological content from documents, such as patents and intellectual property information, is therefore a significant issue in R&D technical knowledge management. This paper introduces an intelligent ontology-based knowledge categorization approach to overcome labor-intensive methods when the number of documents that require analysis exceeds manual processing capacity. The ontology-based document categorization approach requires the use of an artificial neural network (ANN) and pre-constructed ontology schemas for given domains. The system extracts the features of a document by using a morphological analysis and sentence analysis. These features are subsequently matched with classes and relationships of the domain ontology and are transferred as input into the ANN model. The ANN model is trained and tested for the given documents and the assigned categories that are based on the content ontological analysis. Two cases that cover chemical mechanical polishing (CMP) patent documents and IP news clippings are provided to demonstrate the categorization approach for R&D knowledge self-organization.