透過您的圖書館登入
IP:3.144.98.13
  • 學位論文

中文專利指標及文字探勘之研究

Developing Chinese Patent Indexes Extraction and Patent Document Mining Methodology

指導教授 : 張瑞芬

摘要


現今是個知識爆炸的時代,企業為了提升自我競爭優勢,不斷的開發新知識和搜尋舊知識並加以改善。專利文件中的資訊包含許多知識,且這些資訊可以輕易地透過網際網路下載,使用者可快速取得專利文件探討其感興趣的知識。以有許多學者利用各種方法探討專利文件內隱含的知識。利用資料探勘尋找結構化的專利指標資訊,並且將這些資料進行專利分析,探討專利內的技術或預測技術發展的趨勢;也有學者利用文字探勘擷取專利文件內的關鍵字擷,再利用關鍵字進行分析,藉此挖掘專利文件中的知識。 本研究透過自動化的文字探勘及專利分析及專利品質評估,利用斷詞方法將中文句子拆解為單詞,利用系統所拆解出的單詞進行關鍵字擷取並計算詞頻矩陣,利用詞頻矩陣進行技術分群藉此獲得專利的技術分布情形。透過因素分析篩選出關鍵專利指標,利用關鍵指標建立專利品質評估類神經模型,透過建立好的類神經權重可將未知品質的專利進行品質預測。自動化的系統更可提供決策者快速方便的操作介面,在研發新技術或觀察市場技術分布趨勢可提供更快速易懂的資訊,降低決策及技術開發的時間和金錢成本。

並列摘要


In the development of knowledge age, enterprise develops new technologies and improves the existing technology for competitive advantage. The information of patent documents includes a lot of knowledge, and the information is shared on the public patent database. Enterprise can quickly access and study patent content that they are interested. Many researchers use various methods to extract the potential knowledge by analyzing patent documents. They use data mining techniques to find the useful information from patent indicators. Using those indicators analyze the domain technology and forecast the technology trend. Moreover, researchers use the text mining techniques to extract keywords from collected patent documents, and use extracted keywords to find the related patent documents for advance patent analysis. In this research, the proposed Chinese patent evaluation system analyzes the collected patent documents by Chinese text mining and patent quality model. First, the system uses Chinese word segmentation approach to extract Chinese sentences. The system automatic counts the number of Chinese key phrases, and calculates the term frequency matrix from patent documents. Second, the system clusters the patent documents for analyzing the technology groups. In each group, the system uses factor analysis to select key patent indicators. The patent quality model is trained by the extracted patent indicators and the Artificial Neural Network. Thus, the proposed system analyzes the patent technology strength by the trained patent quality model. Finally, the proposed system provides the decision makers with nice user interface and powerful Chinese patent analysis function. Therefore, the proposed methodologies help enterprise in reducing cost and time for research and development.

參考文獻


[1] 方心伶,中文斷詞與注音,碩士論文,清華大學,統計學研究所,2008。
[4] 吳昆璟,以信心量度改善中文斷詞之初探,碩士論文,清華大學,統計學研究所,2008。
[10] 林啟瑋,以主成份分析及類神經網路為基之專利重要性評估方法論,碩士論文,台北科技大學,工業工程與管理系,2010。
[14] 黃盈碩,非耗盡分群方法為基之可重疊專利分群方法論研究,碩士論文,清華大學,工業工程與工程管理學系,2007。
[21] 經濟部智慧財產局,「經濟部智慧財產局-九十八年年報」,2009。

被引用紀錄


蔡志雄(2015)。專利資料探勘方法之研究〔碩士論文,逢甲大學〕。華藝線上圖書館。https://doi.org/10.6341/fcu.M0220238
鍾彧華(2012)。Nichia LED專利策略研究〔碩士論文,國立臺北科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0006-1709201210222400

延伸閱讀