Analysis on integrated Chinese and English term and word segmentation system of the patents of LED

• Main Authors: Wu-sian Lin, 林武憲
• Other Authors: Rong-cing Lin
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/33522380841457681924

碩士 === 國立臺灣科技大學 === 機械工程系 === 98 === Absract The study collects certain patent information, including LED integrated structure manufacturing technique, substrate polishing technique, substrate inspection and testing techniques, and substrate cleaning technique, which are written in traditional Chinese and simplified Chinese, as well as the American patent information written in English. The International Patent Classification (IPC) with related Chinese and English patents established, as well as the term and word segmentation method of integrated Chinese and English patents are both applied to the searched Chinese and English patents mentioned above. The main analytic areas of the study are LED integrated structure manufacturing technique, substrate polishing technique, substrate inspection and testing techniques, and substrate cleaning technique, intending to verify the method established by the study. The established method can also be extensively applied to analysis of other LED-related Chinese and English patents. Having referred to the LED-related engineering knowledge books, together with LED-related patents, the study draws an LED material list, based on which a branch-shaped structural diagram with technical and functional hierarchies is further drawn. Besides, the study establishes an innovative semantic analysis and term segmentation system and method of integrated Chinese and English patents for application to integrated Chinese and English patents. Regarding the new system and method, it applies semantic analysis method of the study to extract the keywords and technical and functional terms of the parts appeared in the Chinese and English patents, and applies the established cross-reference table of traditional Chinese words, simplified Chinese words and English words. The number of times of the appearance of keywords is matched with normalization concept of the number of times so as to establish word clusters. Each of them is provided with around 20 Taiwanese and American patents to train from the bottom to the top the word clusters of different technical and functional hierarchies. Fuzzy values are given and saved in the data list. Then IPC analysis is carried out by computer programs, and a branch-shaped structural diagram of technical and functional hierarchies expressed in IPC codes is drawn. The texts of patents can be searched based on IPC codes, and furthermore, the word clusters can be improved and expanded. After that, LED is taken as the main carrier to analyze different kinds of statistical diagrams of inventors and countries. Focusing on the related patents mentioned above, more detailed analysis is made. A diagram of competitors’ core competitiveness is then established, with companies on column side, IPC codes on row side and the number of patents as the elements. Besides, a main competitors’ R&D activity trend diagram is also established, with companies on column side, years on row side and the number of patents as the elements. Based on the above, patents can be continuously added to expand the word clusters, and increase the accuracy of analysis. Semi-automatically and semi-manually conducting technical and functional analysis of the related patents mentioned above can more rapidly construct a technical/functional matrix of related patents. The established technical/functional matrix can be applied to further establish a technique-based competitors’ core competitiveness diagram. The above searched related patents of LED integrated structure manufacturing technique, substrate polishing technique, substrate inspection and testing techniques and substrate cleaning technique, as well as the analytic results can be provided to enterprises and engineers as a reference in producing new patents, to assist in their researches, save the R&D time and avoid infringement. In addition, enterprises can understand the core competitiveness of competitors.