Computational approach for identifying and visualizing innovation in patent networks
MetadataShow full item record
The importance of intellectual property research has been increasingly addressed in the recent decades. Patent, which is highly structured textual data, is one of the most important types of intellectual property. It's widely accepted that patent documents contain valuable information for industry, business, law and policy making community. Creative solutions, business trends, technological details, and their relationships can be discovered when a thorough analysis is made. However, the gap between current knowledge and the demand from innovators and entrepreneur is still huge. Current computer-aided design (CAD) systems lack capabilities for supporting some design activities, especially during the conceptual design stage. The manual work required for building an ontology library can be time-consuming and error-prone. Besides, there is no dedicating ontology library for support engineering design. Attempts of applying natural language processing techniques to support conceptual design activities are still rudimentary. More transdisciplinary research is desirable in order to move the field forward. Finally, most existing patent analysis methods rely on patent citation data to yield a statistical interpretation. These four reasons motivate this research which aims at satisfying potential requirements from inventors and entrepreneurs who are interested in discovering, improving, realizing, and commercializing new technologies and inventions. The proposed research integrates knowledge from disciplines such as engineering design theories, statistics, linguistics, computer science and software engineering. The successful implementation of the research will enable engineers, inventors and entrepreneurs to analyze the content of up-to-date patents during the conceptual design of new products and the development of new technologies more efficiently. This will spark for more innovations, help avoid patent infringement and thus reinforce the competitiveness of the America. Main scopes in the proposed research includes: a new tree construction algorithm for building patent claim tree based on the ontology library, a comparison algorithm for determining infringement probability between two arbitrary patents, a tree-based patent knowledge representation method and a feature generation method based on the proposed knowledge representation for clustering, an interactive visualization approach which will display clustering results in accordance with user preference, and a software licensed under the GNU General Public License, version 3.0 (GPLv3) that will satisfy all deliverables above with user-friendly graphics user interface (GUI). Data sources of the proposed research include online patent documents retrieved from the United States Patent and Trademark Offices (USPTO), knowledge documented in engineering handbook and dictionaries, natural language processing models from Stanford University, and WordNet from Princeton University. The validation of the proposed methods will be carried out by using the holdout validation scheme and by comparing against manual analysis results.