以複雜網路分析新聞報導中的小世界 －以ECFA相關新聞為例

Abstract

以人物為主的新聞報導中，存在著錯綜複雜的人際網路關係。在探討一個事件的人物關係時，往往會被人物間長久以來所建構的深層關係所影響，而難以找出該事件的關鍵人物。本研究中，利用具有階層架構的複雜網路概念，以階層導向強鍵式及橋接式連結偵測演算法，依據網路的結構與拓譜特性來評判每一個人物關係在各階層可能的強弱性質，並依照這些性質找出人物間既有的黏著性關係及強鍵式關係。透過此方式，得以排除人物間既有的深層關係，找出單一事件中的核心人物。本文以台灣與中國簽訂ECFA的事件中，台灣兩大報紙資料庫為例檢視此模型。透過本研究所提出的演算模型，找出人際關係圖中最重要的關鍵人物為李述德（ECFA事件時的財政部長，為事件主事者），比較以傳統報導次數排名的文字探勘方式找出的前兩名－馬英九及蔡英文（兩個台灣執政黨黨主席）－更貼切本事件的關鍵人物。除此之外，透過模型可提供一套別於傳統比較媒體的方式，以人際網路關係為基礎判斷不同媒體報導同一事件的異同。

Complicated human networks and relationships exist in news reports, which tend to be people-centric. The act of examining the relationship between the characters in a particular event is often influenced by the presence of strong linkages that have been built up between these people over time. As a result, it becomes more difficult to draw out the key characters involved in the event. This research employs the concept of complex networks with a hierarchical architecture, and makes use of the hierarchically-oriented bond and bridge motif detection algorithm method to assess the strength of each relationship at each hierarchical level, based on the structural and spectral properties of the network. These properties are used in turn to extract the existing adhesive relationships and bonds between the characters. Hence, the study of relationships using this method enables the elimination of existing strong linkages between characters, and facilitates the extraction of the core figures involved in a specific event. In this paper, this model is tested using reports on the Taiwan-China Economic Cooperation Framework Agreement (ECFA), obtained from the archives of the two major Taiwanese papers. The proposed algorithm model pulled out Lee Sush-der (Minister of Finance at the time that the ECFA was concluded) as the key figure in the relationship map for this event, a result of greater relevance as compared to the two figures obtained through traditional text-mining methods—Ma Ying-jeou and Tsai Ing-wen (leaders of the ruling political parties).