以創新構思問題解決方法(TRIZ)進行關聯專利檢索之研究探討-以綠色產業為例

A Study of TRIZ for Patent Mapping – A Case Study for Green Industries

游子鋐(Tzu-Hung Yu)

指導教授 ： 葉繼豪

國立臺北科技大學/管理學院/工業工程與管理研究所/碩士(2009年)

https://doi.org/10.6841/NTUT.2009.00468

摘要

創新構思問題解決方法(TRIZ)自蘇聯發明家Genrich Altshuller起始發展至今，漸漸為大眾所熟知，其延伸的相關研究也不斷被提出。目前國內專利的檢索方式及分類方法偏向技術導向，非專業相關技術人員較難從中找到解決使用者問題或提供創新想法的專利內容，因此近年來結合文字探勘及TRIZ方法的專利檢索研究逐漸受到注意，希望提供一種以問題解決基本面為檢索依據的專利檢索方式。鑑於近來世界各國在環保議題的重視、綠色產業的興起，發展產品生命週期各階段均蘊含環保要求的設計日益重要的前提下，本論文所使用專利樣本聚焦在綠色產業，進行TRIZ自動專利分類及建立TRIZ專利檢索機制之研究。收集1000筆綠色產業相關專利，先將此專利集經過特徵產生、特徵萃取、特徵向量化之處理，以支持向量機(SVM)與最近鄰居法(KNN)為分類器，進行分類實驗及績效評估(準確率、精準度、回想率、F(2)-value)；再使用相同的專利集經由字頻(TF)及自動專利分類研究中績效較佳之分類器計算，求得整合39參數、40創新原則與專利間的對應關係，以結合傳統矛盾矩陣及Matrix2003得到之整合39×39 TRIZ矛盾矩陣為基礎，建立一綠色產業之TRIZ高關聯中華民國專利檢索機制。由本論文實驗得到結論，TRIZ自動專利分類中，支持向量機與最近鄰居法的分類預測辨識率差距僅有1%，而評估指標F(2)-value進行評估的結果為最近鄰居法能使TRIZ專利檢索機制具較高可靠度；接著建立綠色產業之TRIZ高關聯中華民國專利檢索機制，利用字頻(TF)與最近鄰居法(KNN)找出整合39參數、40創新原則與綠色產業專利間之關聯性，讓使用者能經由分析問題、找出其中隱藏之取捨矛盾後，透過1組或多組成對整合TRIZ39工程參數之輸入，自動取得TRIZ創新原則之推薦，並且依照工程參數及創新原則由高而低的相關性排列，將高關聯之綠色產業相關專利檢索出來，希望能透過此TRIZ專利檢索機制提供使用者綠色設計、綠色創新的概念至設計發想或問題解決階段。

關鍵字

創新構思問題解決方法(TRIZ) ； 綠色產業 ； 文字探勘 ； 專利分類 ； 專利檢索 ； 支持向量機(SVM) ； 最近鄰居法(KNN)

並列摘要

The inventor of Soviet Union, Genrich Altshuller developed TRIZ to present as well-known to the public. The aim of this research is to map 1,000 Chinese patents of R.O.C Green-related industry into TRIZ Inventive Principles by using text mining technique, and develop a computer-aided mechanism to patents searching by Contradiction Matrix tool and Inventive Principles. There are two parts in the reaserch. First at all, using on-line auto-tag system provided by Academia Sinica to break every sentence in a document into several keywords and label these keywords manually. Secondly, calculating text frequency (TF) and inverse document frequency (IDF) in the corresponding document. Then, chi-square statistics and correlation coefficient approaches are used to select and sort word features which are highly correlated to 40 TRIZ Inventive Principles. In addition, TFIDF and weight-TFIDF values are inputs for further classifiers such as support vector machine (SVM) and k-nearest neighbor classifier (KNN). Finally, SVM and KNN evaluate the performances of 1,000 Chinese R.O.C patents of green-related industry into 40 TRIZ Inventive Principles. Experimental results of part 1 show that, Both of SVM and KNN perform well and the accuracy between them was only 1%. However, used the comparative measure F(2)-value to assess performance of SVM and KNN. KNN could deliver better performance. At part 2, according to text frequency (TF) and KNN build relationships between integrated 39 TRIZ Parameters, 40 TRIZ Inventive Principles and patents. We can build the TRIZ high-correlation R.O.C patents searching mechanism of green-related industry by Integrated 39×39 TRIZ Contradiction Matrix, which combined original classical TRIZ matrix with Matrix2003 developed by E. Domb & M. Slocum. Experimental results of part 2 show that, the user could use the TRIZ high-correlation R.O.C patents searching mechanism of green-related industry to get TRIZ Inventive Principles by analysis of problems to find the hidden contradictions and input one or more pair-wise integrated 39 TRIZ Parameters. Moreover, searching the high-correlation patents of green industry by sorted the correlation of the integrated 39 TRIZ Parameters and 40 Inventive Principles in descending order. We hope that it could provide users with the concept of eco-design and eco-innovation for the phase of product design or problem solving.

並列關鍵字

TRIZ ； Green Industries ； Text Mining ； Patent Classification ； Patent Searching ； Support Vector Machine (SVM) ； K-Nearest neighbor (KNN)

參考文獻

[9] 林秀美，運用TRIZ原理探討專利開發實例，中原大學機械工程學系碩士學位論文，桃園，2004。

[28] 陳冠榮，整合以TRIZ為基之綠色產品開發與生命週期評估系統，國立清華大學工業工程與管理研究所碩士論文，新竹，2007。

[32] 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。

[35] 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。

[44] 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。