透過您的圖書館登入
IP:3.142.35.75
  • 學位論文

萃智(TRIZ)創新高關聯專利檢索與網路平台建置—以中華民國專利與美國專利為例

A Study of TRIZ for Building Patent Searching and Network Platform – A Case Study for Taiwan Patent and America Patent

指導教授 : 葉繼豪

摘要


在這個資訊爆炸、科技日新月異的時代,世界各地的資訊與知識都以無時差地互相交流、分享,使資訊不斷累積、重組再激盪成新的知識。知識經濟儼然成為世界的主流價值。近年來創新能力逐漸超越一個國家的國力與影響力等傳統資源,已成為經濟成長的關鍵。如何有效率地利用可取得的知識資源,透過創新研發來產生具有更大價值的新知識、提升競爭力是我們現在所關心的。 萃智(TRIZ)是一種規則性的創新思考方法,能夠有效率地在短時間內創新、開拓新思維,透過TRIZ創新理論裡面的其中一種工具「矛盾矩陣」,把工程或產品創新上需要解決的問題拆成「提升」和「惡化」兩個維度來思考,分別選擇TRIZ 39工程參數中最相關的其中一項,透過矛盾矩陣所對應的TRIZ 40創新原則來提供使用者解決方案。本研究論文針對兩種新的特徵擷取方法Entropy-TFIDF和Gini-TFIDF特徵擷取方法對1千筆中華民國專利進行特徵擷取與創新原則分類,藉由績效評估結果來探討其分類績效,並建構TRIZ高關聯專利檢索系統,其中包括會員功能、參數選取、自動參數推薦、履歷建置以及專利資料夾等功能;另外使用TRIZ高關聯專利檢索機制搭配擴增的TRIZ 39工程參數與TRIZ 40創新原則之關鍵字表來提高檢索的準確性,以精準度(Precision)和回想率(Recall)兩種績效評估方法來加以探討改良之特徵擷取方法的分類績效。 論文之研究三種特徵擷取方法之分類績效結果顯示,以TFIDF為基礎改良之Entropy-TFIDF特徵擷取方法所獲得的總回想率與總精準度皆優於傳統TFIDF和Gini-TFIDF特徵擷取方法,反觀Gini-TFIDF特徵擷取方法之分類績效結果則不如預期,總回想率與總精準度皆低於傳統TFIDF。在TRIZ高關聯專利檢索系統的實作上,不只擴增了中華民國專利資料庫也首次將TRIZ創新構思解決方法與美國專利資料檢索結合在一起,並透過TRIZ 39工程參數與TRIZ 40創新原則之組合從中華民國專利與美國專利各一萬筆專利中檢索出具有高關聯的專利資料,透過修正回想率和修正精準度加以評估檢索績效之後發現,對於某些參數-原則組合之專利檢索具有一定的績效水準,而使用者還可以透過調整權重門檻值來檢索出符合需求的國內外專利資料,當權重門檻值設定較高時,可以檢索出關聯性高且較聚焦的專利資料,反之,權重門檻值設定較低時,則可以檢索出範圍較廣且數量較多的專利資料,此TRIZ高關聯檢索系統還提供了專利儲存功能,提供了使用者更方便、更彈性的檢索方式。

並列摘要


In this versatile era, information and knowledge are sharing and transmitting in sync, which allows information to be accumulated and then recombined as new information. Knowledge-driven economy becomes the main branch in this world. The ability to innovate is the key to economic growth in recent years. How to obtain resource efficiently and promote competitive strength through creation thus becomes a matter of utmost importance. TRIZ is a rule of innovative thinking method which efficiently in a short period of innovation to explore new ideas. TRIZ contradiction Matrix is able to split the problems into two dimensions - upgrading and deterioration. We can solve the contradiction problem by using TRIZ 39 parameters, TRIZ 40 innovative principles and mapping from TRIZ contradiction Matrix. The aim of this research is to map 1,000 Taiwan traditional Chinese patents into 40 TRIZ innovative principles by using two new text mining algorithms, Entropy-TFIDF and Gini-TFIDF, which are developed from traditional TFIDF. This research evaluates the classification performance by Entropy-TFIDF and Gini-TFIDF by using indices such as precision and recall. It provides TRIZ users a basic idea to search more correlative patents effectively for solving difficult problems by inspired by TRIZ 40 innovative principles. The results show that Entropy-TFIDF outperforms Gini-TFIDF referred the classified by TRIZ 40 innovative principles no matter in precision or recall. And TRIZ Patent Searching System which involves 10,000 ROC patents and 10,000 USA patents is highly associated with good retrieval performance. The retrieval performance in this system can achieve at the recall rate of 87.5% and precision of 61.5%. Therefore, TRIZ Patent Searching System is very useful for design thinking and product innovation. Finally it is hope that it could provide users with the concept of eco-design and eco-innovation for the phase of product design or problem solving.

參考文獻


[12] 林秀美,運用TRIZ原理探討專利開發實例,碩士論文,中原大學機械工程研究所,桃園,2004。
[17] 顏吉承、陳重任,設計專利-理論與實務,台北:揚智文化事業股份有限公司,2007。
[19] 經濟部智慧財產局,認識專利,台北:經濟部智慧財產局,2008。
[23] 經濟部智慧財產局,專利法施行細則,台北:經濟部智慧財產局,2002。
[25] 陳稼興、謝佳倫、許芳誠,以遺傳演算法為基礎的中文斷詞研究,電子商務學報,第二卷,第二期,2000,第27-44頁。

延伸閱讀