電子目錄系統個人化商品資訊過濾之研究

Abstract

網際網路的發達使得越來越多的虛擬商店在網路上販賣商品，然而，過度氾濫的資訊使得消費者必須花費相當多的時間去個別搜尋自己所感興趣的商品。電子目錄系統可以提供商品資訊的搜尋，但是，仍有可能因為商品資訊過於繁多而讓搜尋結果過於龐大。 本研究之目的在於發展一架構在電子目錄下的個人化商品資訊過濾方法，從電子目錄進一步去過濾商品資訊。本文提出一商品資訊的描述方式，來表達商品的特性，並建立三層的商品樹來描述商品間屬性的樹狀關係，並根據電子目錄的架構，建立出使用者對於不同商品特性的三層式偏好設定，來描述使用者購物的喜好。另外，本文提出反向索引結構來表達商品屬性的關係。最後，本研究提出了根據使用者偏好設定分別針對商品樹及反向索引結構進行搜尋與過濾的不同方法，並比較其在資訊過濾的效能。

With the rapid growth of Internet, virtual stores are increasing enormously. However, users may encounter an information over-flooding problem, which makes users spend most of their time searching and eliminating irrelevant merchandise information. Though e-catalog systems provide some search functions to search merchandise information, the result may still be too large. The aim of our study was to design information filtering functions in e-catalog systems that are able to eliminate irrelevant merchandise information automatically by taking into account of user profiles. A three-level merchandise metadata is used to describe the detail of a specific product and to construct the e-catalogs. Additionally, we use a three-level merchandise tree to categorize merchandise information into a tree structure based on the e-catalogs. A three-level user profile based on e-catalog systems is also proposed to describe a user's shopping interest in all kinds of product. Furthermore, an inverted index structure is also used to provide an alternative way to describe the merchandise information. Finally, we propose different matching algorithms to filter merchandise information according to the user profile. Experimental evaluations are conducted to compare the performance of various approaches.