專利探勘之研究：專利分類與專利趨勢變化探勘

Abstract

競爭情報有助於企業經營者判定宏觀環境的利基，在企業決策上扮演著相當關鍵的角色。無論是在企業範疇或是總體層面上分析競爭情報，專利資訊絕對是一種衡量企業競爭能力的重要依據。本論文針對不同的專利管理目標提出兩種不同的方法：1)複合式專利分類以達成專利資訊自動化分類及2)專利趨勢變化探勘以偵測專利研發行為的變化趨勢。 為能更精確的進行專利文件分類，本研究提出的複合式專利分類方法除了整合傳統之內文式、連結式及銓敘資料式專利分類方法外，還包含本研究所提出之專利網路式分類方法。此專利網路除了包含專利文件外，也涵蓋了由專利文件中取得的各種不同特徵作為節點，節點間的連結關係則得自於專利銓敘資料。專利網路式分類法透過分析專利網路中所有可達節點以計算與欲分類專利文件的相關度，並將相關度高的節點作為專利分類的依據。本研究同時也提出一個改良式的k-nearest neighbor分類器以作為分類之用。我們以從美國專利局(United States Patent and Trademark Office)所收集的專利文件做為測試資料，以評估本研究所提之專利網路式分類及複合式專利分類方法的效能。實驗結果顯示專利網路式分類及複合式專利分類方法皆優於傳統的專利分類方式，其中複合式專利分類方法也優於專利網路式分類方法。 本研究所提出之專利趨勢變化探勘方法可以在不需要專業知識的情況下找出隱含在專利資料中的趨勢變化，可分為專利收集、專利指標計算、及變化探勘三個步驟。在變化探勘階段，本方法將從不同時間區段專利資訊中挖掘出趨勢(以rule呈現)，再比較不同時期的專利趨勢以找出趨勢變化，根據趨勢變化的方式可以分成四個種類並分別計算出變化程度，最後將變化程度多寡排序後提供給管理者做為決策之用。我們將專利趨勢變化探勘方法用於台灣的半導體產業分析上，以找出四種不同層級的專利趨勢:競爭對手的研發行為變化、產業領導者在特定技術領域的研發行為變化、產業領導者研發行為變化及產業特定技術領域的趨勢變化。

Before executives or mangers make strategic decisions for an organization, competitive intelligence always plays a critical part on identifying niches within macro environment. For analyzing the competitive intelligence, either within a business scope or on a global view, patent is absolutely the most visible factor on evaluating competence of each participating business. This thesis proposes two approaches for different patent management purpose: the hybrid patent classification approach for automatically classifying patents, and the patent trend change mining approach for detecting technological change trends. The hybrid patent classification procedure combines classic content-based, citation-based and metadata-based methods, with a novel patent network-based method to perform patent classification. The proposed patent network, which contains various types of nodes that represent different features extracted from patent documents, and the nodes are connected based on the relationship metrics derived from patent metadata. The novel approach analyzes reachable nodes in the patent ontology network to calculate their relevance to query patent, after which it uses the k-nearest neighbor classifier to classify query patents. To further improve the approach, it is combined with content-based, citation-based and metadata-based classification methods as the proposed hybrid classification approach. We evaluate the performance of the hybrid approach on a test dataset of patent documents obtained from the United States Patent and Trademark Office (USPTO), and compare it with the performance of the three conventional methods. The results demonstrate that the proposed patent network-based approach outperforms the conventional approaches, and the proposed hybrid classification approach performs better than the patent network-based approach. The proposed patent trend change mining (PTCM) approach can identify changes in patent trends without the need for specialist knowledge. The proposed approach consists of steps including patent collection, patent indicator calculation, and change detection. In change detection phase, the approach firstly extract rules between two different time periods, comparing them to determine the trend changes. These trend changes are then classified into four categories of change, evaluated with change degree and ranked by their change degree as the output information to be referred by decision makers. We apply the PTCM approach to Taiwan’s semiconductor industry to discover changes in four types of patent trends: the R&D activities of a company, the R&D activities of the industry, company activities in the industry and industry activities generally. The proposed approach generates competitive intelligence to help managers develop appropriate business strategies.