部落格內連結定義對小世界網路之影響

Abstract

隨著社會網路網站(SNS)的興起，網際網路越來越趨向社會化，有更多的研究著重於利用社會網路特性來改善資訊搜尋及擴散效能，但是這些研究忽略了影響網路結構的連結定義，如同真實社會的複雜關係，社會網站內也存在許多複雜多樣的連結，不同的連結定義可自成一格形成一種網路架構，然而，網路特性將隨著架構改變；在許多社會網路的特性中，冪次法則分配、高群聚係數及短平均路徑合稱為小世界網路的三大特性，本研究將深入探討連結定義對小世界網路的影響，從另一個角度提升社會網路中的搜尋及擴散效能。 本研究以天空部落為資料來源，分別擷取「好友」及文章中的「留言」做為連結，建構出兩種具方向性的網路；另外篩選出「嚴格好友」、「寬鬆好友」及「嚴格留言」的連結定義建構出三種無方向性的網路；我們發現即使連結定義不同，這些網路依舊屬於小世界網路，其中「留言」定義總連結數並不如想像密集，但卻有和「好友」相當的平均路徑長；「寬鬆好友」定義擁有最短的平均距離；「嚴格好友」具備最好的網路容錯性。最後實際比較出這五種連結定義在資訊搜尋及擴散效能的優劣，發現擴散效率上寬鬆好友>嚴格好友>嚴格留言，搜尋效率上寬鬆好友>嚴格好友>好友>嚴格留言>留言。

Social network sites (SNS) have been springing up during the last decade, and because of so, the human interactive behavior started to happen on the Internet. A number of recent studies have focused on using specific properties of social network to upgrade information diffusion or searching performance, but they neglected that there are miscellaneous relations existing in SNS just like the real world, namely, there are more than one link definition to be chosen as the network edges. Since network is composed of links, apparently different kind of link could alter network properties. Our research goal is to thoroughly observe the impact of each link definition on Small-World network. Furthermore, we demonstrate some simple diffusing and searching strategies on different topology to see the diversity of efficient performance. We collected data from Taiwan’s online social network—Yam blog, exploring along link “Friend” and “Comment”, then filtering out some nodes to obtain “Friend(strict)”, “Friend(loose)” and “Comment(strict)” graphs. We find that they still have the Small-World property. Besides, “Comment” topology is less dense than we thought, but still has a short average distance near “Friend”; “Friend(loose)” has the shortest average distance; “Friend(strict)” has the best network resilience. In the end, we compare them by using one diffusing and two searching strategies. The ranking of diffusing efficiency is Friend(loose)>Friend(strict)>Comment(strict), while for searching is Friend(loose)>Friend(strict)>Friend>Comment(strict)>Comment.