請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60910
標題: | 偵測科技機會:藉由語意與引文分析衡量專利新穎性 A New Approach for Technological Opportunity Identification: Patent Novelty Assessment by Semantic and Citation Analysis |
作者: | Guan-Yu Pan 潘冠宇 |
指導教授: | 魏志平 |
關鍵字: | 專利分析,文字探勘,自然語言處理,專利新穎性, Patent Analysis,Text Mining,NLP,Patent Novelty, |
出版年 : | 2013 |
學位: | 碩士 |
摘要: | 在現在這個海量資訊充斥的大環境中,專利管理一直是科技密集產業最為需要重視的課題,企業不僅僅要透過專利來掌握並保護公司的特有科技,還要去了解到其他競爭者在同產業的科技發展與專利布局狀況,透過智慧財產戰爭來獲得競爭優勢已經非常常見。然而根據資訊溢出理論(theory of information overflow),現在要透過人工來處理這些專利的大量資料變得相當困難,不僅費時費人力,還要付出大筆資金才能完成分析工作。而本論文將會討論過去與科技機會相關的文獻,並建構出一個新的方法來偵測科技機會。主要理論來自於文獻中的語意分析及引文分析專利新穎性。實驗結果顯示我們提出的方法,經由訓練後所得之最佳新穎性分類器(C4.5決策樹演算法),可以達到F1分數76.6%及AUC分數71.8%,顯示了良好的新穎性預測能力。 In the day with a huge information and high technology, the patent management issue is an important factor for technological-intense firms. These companies not only need to hold the exclusive technology to capture the R&D competence but have to know the whole technological field to fight for intelligence property war. However, according to theory of information overflow, it is hard to analysis from vast data because of expensive manual cost and time-consuming tasks. In this paper, we collect and review the previous studies and propose a new approach to discover technological opportunities, which is using patent novelty measurement by a hybrid framework of semantic and citation analysis. Otherwise, it also can help relative applications development and enhance follow-up studies for technological opportunities identification. The result of our experiment shows that the best novelty classifier (based on C4.5 decision tree) can reach 76.6% on F1-score and 71.8% on AUC. That shows a good performance of novelty prediction. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60910 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊管理學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-102-1.pdf 目前未授權公開取用 | 473.07 kB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。