| Summary: | 碩士 === 國立臺灣科技大學 === 機械工程系 === 106 === With urgent demand on electronic power system of high power performance, mono-crystalline Silicon carbide wafers have been considered as a high potential materials for high power IC devices. However, the ultra high hardness and excellent chemical stability of SiC wafers induce a very long processing time duration and high cost in chemical mechanical polishing (CMP) process. Many researches have indicated some potential improvement of CMP process of SiC wafers, a roadmap of future development is definitely needed to achieve above challenges. Patent deployment plays an important role in research resource, but it is inefficiently if all operations depend on manual work. This study aims to use text mining and ontology-based technology to enhance the efficiency of patent analysis. This study has been searched through Orbit patent database and collect 837 global patent families of SiC wafer chemical mechanical polishing (SiC CMP) from 1986 to 2017 and also analyzed by R language programming. Pre-constructed ontology method is used for patent classification, technical concept and function analysis. In patent classification, ontology phrase is an index that used to classify various technical patents and also applying the Pearson's correlation coefficient to extract the CMP related ontology phrases. In technical concept and function analysis, phrases with highest TF-IDF frequency are selected as results which based on specific rule of key phrases selecting. Results of this study can develop and obtain the classified road map and technique-function matrix. Results of evaluation rate of patent classification precision is 93.93%, concept precision is 91.85% and function precision is 84.25%. Future study can focus on developing a smart patent analysis system.
|
|---|
