在JXTA同儕網路上建構關鍵字搜尋服務

Abstract

JXTA(Juxtapose)是一組傳輸協定，可提供任何可以上網的設備在同儕網路(P2P)架構下進行傳遞訊息及協同運作。JXTA具有成為下一世代同儕網路平台的潛力，並且已被應用於即時通訊、檔案分享及即時合作平台等應用程式。但是JXTA缺乏關鍵字搜尋的功能，導致查詢(query)必需使用搜尋對象的全名才有辦法找到。這樣的搜尋方式對使用者是很不方便的。本篇論文提出了一個JXTA關鍵字搜尋(JKS)服務並且支援中文關鍵字切割(CKP)。JKS 有兩個目標，第一個目標是設計新的發佈(publishing)和搜尋機制(discovery)，以達成JXTA關鍵字搜尋服務。第二個目標是設計中文關鍵字切割方法，以提升完全符合查詢條件的搜尋結果數量(不含重覆)。根據一些知名入口網站資料所做的模擬結果顯示，當資料大部份為中文時，JKS較KAD增加了427%完全符合查詢條件的搜尋結果數量。在網路流量方面，當資料全為英文時，JKS較KAD少70%。當資料主要為中文時，JKS較KAD少11%。評估結果顯示JKS能夠應用於JXTA上的應用軟體，諸如檔案分享、多媒體串流分享及網路服務搜尋等。

JXTA is a set of protocols which provide different network devices to communicate and collaborate in a P2P (peer-to peer) manner. It is emerging as the next-generation P2P platform and has been adopted by many applications, such as instant messaging systems, file sharing systems, and real-time collaboration platforms, etc. However, JXTA is lack of keyword search. Without the support of keyword search, the query must contain the exact full name of a desired resource. It is inconvenient for users in this aspect. This thesis proposes a mechanism called JXTA Keyword Search (JKS) with the support of Chinese Keyword Partition (CKP). JKS has two objectives. First, it is providing keyword search upon JXTA. New publishing and discovery schemes are proposed for keyword search. Second, it is designing a Chinese Keyword Partition method to enhance the number of exact matches. Experimental results, based on real data obtained from some well-known portal sites, show that the number of exact matches of JKS is 81% larger than that of KAD when the majorities of resources and queries are Chinese. JKS is 237% and 12% less than KAD in terms of bandwidth cost when resources and queries are all English and the majorities of resources and queries are Chinese, respectively. The research results are applicable for P2P applications (e.g., file sharing, multimedia streaming sharing, internet service discovery, etc.) built on JXTA.