專利標籤網路之動態分析

Abstract

隨著連結時代的到來與網路科學的興起，探討各種關係所構成的網路組織以及分析其結構特性與應用的相關研究日益受到重視，例如人際關係所構成的疾病傳播網路、人際交友網路、資訊所賴以傳播的電信與 E-mail網路、或是實體關係所構成的電力輸送網路、鐵路網路等，文件之間藉由特殊關係所構成的文件網路也是網路科學所探討的對象之一，例如網路文件間藉由相同的使用者標籤所構成的知識網路。 專利文件利用彼此之間引證的關係可產生專利引證文件網路，然而專利引證文件網路的建立涉及自然語言處理技術，本身有其技術門檻，因此本研究採新的思維，採用其他方式建構專利文件網路並實際應用於網路分析，試從日本專利局的多維度專利分類系統 (Multi-Dimensional Classification System )中技術特徵碼和專利之間的關係建構出專利標籤網路，並且利用專利標籤網路進行動態分析。網路分析的項目分為（ 1）標籤網路的網路基本特性（ 2）專利單元網路的動態分析（3 ）技術特徵碼網路的動態分析。 經由網路分析，除了能夠進一步了解 "網路基本性質" 之外，還能夠對"專利網路隨時間變化 "、" 產業技術發展趨勢"、 "以專利為中心的關聯網路" 及"以技術特徵為中心的關聯網路 "作進一步的探討，另就應用面而言，還可藉由動態分析得到重要的專利趨勢變化與競爭情報。

According to the rising of Network Science, the researches related to this field are growing up progressively. For example, the epidemic broadcasting networks or friend networks which connected by interpersonal relationship, telecommunication networks or e-mail networks which connected by information circulation, power line networks or railway networks which constructed by physical relationship. The document networks which connected by the logical associated text data such as the knowledge networks in the cyberspace by the user tagging and patent citation networks are also the research objects of network science. Patent citation networks are generated by the relationship of citing between patent to patent, however the connecting need the Natural Language Processing (NLP), and the NLP is the major difficulty of patent network connecting, thus in this thesis we adopt a new approach, using another method to construct patent document networks, we use the connecting relationship between technical codes and patents from the Multi-Dimensional Patent Classification System which made by Japanese Patent Office (JPO) to construct the Patent Tagging Networks and using the patent tagging networks to analyze this network. The network analysis can divide into (1) The fundamental characters of the tagging networks (2) The dynamic analysis of the patent unit networks (3) The dynamic analysis of the technical code unit networks. In the thesis, we concern about how to construct a Patent Tagging Networks and then we focus on the application of the Patent Tagging Networks, thus the dynamic network analysis.