異質網路環境之行動搜尋關鍵技術---子計畫五：行動搜尋資料萃取與探勘機制

Abstract

隨著無線傳輸技術的發展，與各式各樣無線手持裝置的進步，無線手持裝 置可利用無線網路互相連結，因此，使用者可以透過無線手持裝置進行傳網際網 路的搜尋服務，尋找所需要的資訊，此類型的搜尋服務稱為行動搜尋。手持式裝 置的有限資源，包含了記憶容量小、顯示介面小、輸入不易等，因此，如何設計 與研發一行動搜尋平台，克服行動手持裝置的限制，讓使用者願意使用行動搜尋 的平台，將是一極為重要的研究課題。 本計畫為一為期三年之總計畫『異質無線網路中行動搜尋關鍵技術研發』 之一子計畫。在此子計畫中，我們將研發於本子計畫的主要的研究目標將提出一 個整合性的行動搜尋服務平台，有鑑於 Web 2.0 有用得資訊與豐富的網頁知識， 我們將結合Web2.0及網路上的社群結構(community)，提供使用者有用的資料，並 將研發資料探勘技術，分析使用者的行為模式，輔助預測其欲搜尋的項目，藉以 減少使用者需要輸入的關鍵字及降低因修改關鍵字而重覆查尋的次數。本計畫 中，我們主要的研究課題為 (1).從 Web 2.0 網站進行一資料的重整，將地理物 件如餐廳或是景點，進行資料淬取，以整理出結構化之資料形式，提供行動搜尋 之內容。 (2).針對 Web 2.0 網站，進行社群探勘，以便提供行動搜尋更為有用 且精準的資訊。(3).研發使用者行為探勘機制，以主動提供資料，將行動搜尋得 整體使用效率達到最佳化。 隨著無線網路與手持式裝置的進步，我們相信此計畫之執行，將可研發出 行動搜尋的資料處理前瞻性技術，帶動無線網路應用服務之發展。

Due to the recent advance in embedded techniques and wireless networks, mobile devices with powerful computation, comparatively large storage and wireless communication capabilities are becoming increasingly popular. With these powerful mobile devices, users are able to search a variety of information that are highly interested by users. Such a search service in mobile computing environments is called a mobile search. Although there already exist some mobile search services, most of these mobile search services are directly adopted from services for desktop PCs without taking the constraints of mobile devices into consideration. These resource constraints on mobile devices call for the design of mobile search platform. This project is under the integrated project 『Key Technologies for Mobile Search in Heterogeneous Networks』, and aims at designing information extraction and mining mechanisms for mobile search. Our primary goals include (1). discover Geo-Entity (referred to Geographical Entity) from a huge Web 2.0 repository. (2). mine popular co-cited communities in a blogspace. (3). develop mining algorithms to mine sequential patterns across multiple attributes for overcoming the inconvenient of input in mobile devices. More specifically, with the fast growth of Web 2.0 services, there Web 2.0 information is very useful for mobile search. Thus, we will first propose algorithms to extract Geo-Entity which is related to geographical entity such as a restaurant and sightseeing. Once Geo-Entities are extracted, we will further to mine community structures in Web 2.0. In light of the community mined, a community-based recommendation for ranking is developed. In addition, the logs of user experiences in mobile search will be generated. These logs contain locations, query terms and clicking behaviors when users conduct searches via their mobile devices. With sufficient attributes in the logs, we will further develop mining algorithms to mine sequential patterns across multiple attributes. These sequential patterns are able to provide more in-depth information of user behaviors in mobile search. These techniques will be integrated under smart phones or PDAs and we will evaluate such mobile search techniques by building one prototype mobile search. In view of the increasing attention on mobile search, we strongly believe this project is very timely and will deliver results of both theoretical and practical importance.