dc.description.abstract | Technology intelligence plays a crucial role for corporate strategy, especially during times of accelerating technological progress. Accompanying advancements in computing capacities and natural language processing tools give today’s decision-makers access to a broad range of market information. Particular deep insights are made possible through patent data, for example by measuring the similarity between individual documents, or whole IP portfolios.
Progress in the literature that focuses on accessing this wealth of information, termed IP analytics, is often mis-aligned with the requirements at a company-level: most research either use patent data as an exemplary case to showcase new algorithms for big (textual) data analytics, or they focus on insight-generation for policymakers and other researchers, leaving company decision-makers without practical, applicable solutions.
Therefore, the main contribution of this thesis is a practical and re-applicable framework to assess patent similarity via semantic and categorical means, combined into a hybrid model that can be used on a given portfolio of US patents. The outcome of this model is a weighted list of patent assignee organizations, which are strategically relevant to the initial portfolio.
Through a case study of the US patent portfolio of a medium-sized German firm, it is shown that the proposed hybrid similarity framework can automatically and accurately identify relevant market players, enabling company decision-makers with technology intelligence in a clear and concise way. Both measures of patent similarity were shown to be positively correlated with the strategic importance of the identified assignees to the target company.
All R code developed is available for replication, and application on new patent portfolios at: https://github.com/janikweigel/IP_Similarity_Thesis | en_US |