Abstract
Nowadays, the search for relevant patents plays an important role in product life cycle. If a product manager can retrieve relevant patents according to his/her criteria for a new product quickly and accurately, s/he will be able to efficiently and effectively judge if the tentative product has the complete feature set, any risks in patent infringement, and its potential in technology competition. However, the increasing number of patents creates the additional efforts in patent search in terms of quantity and complexity. Bewaring the emerging need of prior art retrieval from a large patent database, the main objective of this study is to develop a methodology which can build the claim hierarchy automatically in order to retrieve the most relevant patents for a product with technologies from various disciplines. Most existing patent retrieval tools use keywords to find relevant patents; however, they ignore the structure of the invention, which mentions in the claim of patent. Thus, this study parses the sentences of a patent claim and finds the structure of the invention. Then, we compare the similarity of the structure of inventions and retrieve the most relevant patents. In order not to ignore potential relevant patents, the proposed relevant patent retrieval method uses keyword-based patent retrieval technique first. Subsequently, the system uses the automated generated claim trees to retrieve the most relevant patents from the patent set retrieved by keywords. Finally, users could adjust the inner and core patent set to decide how much relevant patents retained as the outputs. The major contribution of this thesis is to decrease the effort for product managers to search relevant patents in product development life cycle. This thesis can serve as a benchmark for researchers to compare with follow-up tech mining techniques.