مقاله انگلیسی رایگان در مورد استفاده از داده کاوی اطلاعات برای بهبود خدمات کتابخانه دیجیتالی – امرالد 2010

Introduction

Although we are today overwhelmed with data, techniques for finding appropriate information are mostly based on syntax search or low-level multimedia features. For improving search results in interaction with digital libraries (DLs), other more intelligent techniques should be used, based on both top-down knowledge creation (e.g. ontologies, user modeling) and bottom-up automated knowledge extraction (e.g. data mining, web mining) (Chen, 2003).

Valuable information extracted from the collection of DL data can be integrated into the library’s strategy, and can be used to improve library search (Chang and Chen, 2006). For an effective design of systems and particularly to help users to find information more easily, it is crucial to understand how people perform searches. This is especially important in continuous development of technologies. By exploring users’ behavior we try to understand better the users themselves and their information needs and provide them with better user-oriented applications. To achieve this goal, we can anticipate a specific user’s needs and problems in advance, by using experience of other similar users.

The idea of a recommender system is to help users by advising them on relevant products/information by predicting in advance their interest in a product; this prediction is based on various types of information, e.g. users’ past purchases and product features (Huang et al., 2002). For a DL, user recommendations may be very helpful (Geisler et al., 2001, Liao et al., 2009).

To help DL users obtain useful information more easily, we can use data mining techniques. Since data mining techniques are very popular, many researchers have applied them in various domains. However, few are focused on the domain of DLs. Our main objective is to use data mining techniques to recommend specific services to DL users.

We have developed the REKOB system to support the users of a specific digital library called KOBSON (http://kobson.nb.rs). In REKOB, we apply different and efficient data mining techniques for clustering DL users based on their profiles and their search behavior. We do not apply data mining to the library documents, but to its services; thereafter we recommend an appropriate service to a new user. By services, we assume online journal services such as Science Direct, Springer, and Blackwell among others.

The paper is organized as follows: a review of related work is discussed in section 2; section 3 describes the REKOB system architecture; experimental results and evaluation are provided in section 4; and finally, we draw our conclusions and plans for future work in section 5.