專利文獻記載了全球90%的技術成果，記載的技術受到各國專利法的保護，但隨著世界技術競爭日益激烈，各國企業紛紛展開專利的戰略研究，因此在專利的分析和運用就受到了企業的重視，專利分析是針對專利說明書和專利文件中大量的訊息內容進行分析、加工、組合並利用統計、資料探勘(Data-mining)、文本挖掘(Text-mining)技巧使這些信息轉換成能幫助企業進行決策、預測的競爭情報，因此專利分析成為企業永續生存和保護商業技術的武器之一，在過去專利分析上針對趨勢分析的研究大都以統計分析的方式針對關鍵字的數量和專利數量進行預測分析，但所能找出的關鍵字(keyword)都侷限於已然成熟的技術並無法找出隱含的新興字詞，因此過去的專利分析都只能找到明顯且具有重要性的字詞，但並未能找到不明顯但對未來技術有重要影響的新興字詞，因此如何找出這些低頻性質的字詞做出正確的趨勢預測是非常重要的研究議題。
本研究採用中文斷詞系統找尋專利文件的字詞，根據Cross-Collection Mixture Model的機率模型來萃取字詞，此模型將針對字詞在時間序列的變化之下，藉由模型中background model及common theme去除掉過於頻繁且不具有分辨意義的字詞和收集在時間變化之下持續出現的字詞，此方法可以快速且大量地篩選專利文件，並且從專利摘要萃取出具有低頻性質的新興字詞，此方法可以順利的篩選掉熱門字詞並且準確的從專利文件偵測出新興技術(emerging technology)的未來趨勢。

Patent has recorded over 90% of the technique worldwide, patent has also been protected by the law in each country. However, as the technology completion has risen up nowadays, the business in each country has started the patent war, therefore, the analysis and implementation of patent has became more important in every business. Patent analysis is focusing on analyzing and combining the message from patent documentations. With statistics, data mining, and text mining, the message can be transformed into a huge role in decisions making and future predictions. Therefore, patent analysis has become a weapon for business to survive and protect their technology. In the past, the majority of the research in trend analysis uses statistics analysis to analyze the amount of keywords and patents. However, the keywords that could be found are limited in the technique that has been developed in years and no more new words could be found. And due to patent documents has the necessity to unveil the technique, the business uses substitute words or phrases to avoid the new words been found. Therefore, patent analysis can only find some obvious and important words but not the key words.
This research use Chinese break words system to find the key word in patent documents, and based on Cross-Collection Mixture Model’s probability model to pick the words. This model uses the time sequences difference of the words, and uses the background model and common theme to delete frequent and indistinguishable word and common theme to collect the words the keep appearing under times. The patent documents can be quickly filtered and found the low appearing frequency and distinguishable words due to automation. Therefore, the searching and filter the popular but aged technology, and precisely detect the emerging technology from patent documents.

[1] A.L. Porter, A.T. Roper, T.W. Mason, F.A. Rossini, J. Banks, Forecasting and Management of Technology, Wiley, New York, 1991.
[2] Murat Bengisu T, Ramzi Nekhili ,Forecasting emerging technologies with the aid of science and technology databases (2005)
[3] Tugrul U. Daim , Guillermo Rueda, Hilary Martin, Pisek Gerdsri, Forecasting emerging technologies: Use of bibliometrics and patent analysis, Technological Forecasting & Social Change,(2005)
[4]Byungun Yoon, Yongtae Park, A text-mining-based patent network: Analytical tool for high-technology trend, Journal of High Technology
Management Research, (2003)
[5] Young Gil Kim, Jong Hwan Suh, Sang Chan Park, Visualization of patent analysis for emerging technology, Expert Systems with Applications, (2007)
[6] Jong Hwan Suh, Sang Chan Park , Service-oriented Technology Roadmap (SoTRM) using patent map for R&D strategy of service industry,(2008)
[7] Sungjoo Lee, Seonghoon Lee, Hyeonju Seol , Yongtae Park, Using patent information for designing new product and technology: keyword based technology roadmapping, R&D Management ,(2008)
[8] Sungjoo Lee,ByungunYoon,YongtaePark, An approach to discovering new technology opportunities: Keyword-based patent map approach, Technovation 29(2008)
[9]Satoshi Morinaga , Kenji Yamanishi, Tracking Dynamics of Topic Trends Using a Finite Mixture Model, ACM 1-58113-888-1/04/0008,(2004)
[10]Yonghui Wu , Yuxin Ding , Xiaolong Wang , Jun Xu , Topic Detection by Topic Model Induced Distance Using Biased Initiation, ADVANCES IN COMPUTER SCIENCE AND INFORMATION TECHNOLOGY (2010)
[11]Qiaozhu Mei, ChengXiang Zhai, A Mixture Model for Contextual Text Mining, ACM 1-59593-339-5/06/0008 ,(2006)
[12] A.L. Porter, A.T. Roper, T.W. Mason, F.A. Rossini, J. Banks, Forecasting and Management of Technology, Wiley, New York, (1991).
[13]J. Perkiom ,W. Buntine, and S. Perttu. Exploring independent trends in a topic-based search engine,(2004)
[14]C. C. Chen, M. C. Chen, and M.-S Chen. Liped:Hmm-based life profiles for adaptive event detection,(2005)
[15]期望值最大演算法, http://ccckmit.wikidot.com/st:em1
[16]webpat台灣 , http://webpat.twipr.com.ezproxy.lib.ncu.edu.tw/WEBPAT/WebpatDefault.aspx
[17]科技產業資訊室,http://cdnet.stpi.org.tw/techroom.htm
[18] Byungun Yoon, Yongtae Park,A text-mining-based patent network: Analytical tool for high-technology trend, Journal of High Technology Management Research,(2003)
[19]Tsai Yu-Fang and Keh-Jiann Chen,”Reliable and Cost-Effective Pos-Tagging”, International Journal of Computational Linguistics & Chinese Language Processing, (2004)
[20] Peng Zang, ChengXiang Zhai,CTMS: A Comparative Text Mining System,
[21] ChengXiang Zhai, Atulya Velivelli, Bei Yu, A cross-collection mixture
model for comparative text mining, ACM 1-58113-888-1/04/0008,(2004)
[22]蔡明誠,專利法,經濟部智慧財產局,(民國95)
[23]賴文智,智慧財產權契約,經濟部智慧財產局,(民國96)
[24]李駿翔,應用資料探勘分類技術於專利分析之研究,中原大學,(民國92)
[25]曾元顯,專利文字之知識探勘：技術與挑戰, 現代資訊組織與檢索研討會,(2004)
[24]經濟部 智慧財產局
[26]曾陳明汝、蔡文誠，兩岸暨歐美專利法，(2009)
[27]童國恩,利用機會探索理論偵測新興專利技術,真理大學管理科學研究所,(民國96)
[28]黨情娜,專利分析方法和主要指標,(2005)
[29]洪秉儒,動態主題截取在網路文件分群之應用,國立中央大學企管所,(2011)