Acquisition, representation and management of user knowledge Alejandro Peña Ayala * WOLNM, 31 julio 1859, # 1099 B, Leyes Reforma, Iztapalapa, DF 09310, Mexico ESIME, National Polytechnic Institute, Unidad Profesional ‘‘Adolfo Lopez Mateos”, edificio 5, 3er piso, La Escalara, Gustavo A. Madero, DF 07738, Mexico Centre for Computer Research, National Polytechnic Institute, Av. Juan de Dios Batiz s/n esq. Miguel Othon Mendizabal, Unidad Profesional ‘‘Adolfo Lopez Mateos”, Zacatenco, Gustavo A. Madero, DF 07738, Mexico article info Keywords: Web-based Intelligent Systems (WBIS) Web-based Education System (WBES) User knowledge Cognitive maps Fuzzy-causal inference abstract Web-based Intelligent Systems (WBIS), e.g. information retrieval, intelligent Web, and e-Learning, deal with tasks such as acquisition, representation, and management of knowledge about users. Based on a user profile, WBIS are able to behave according the particular needs of people through the intelligent adaptation of services, content, navigation interfaces, and many more factors. Thereby, the design of an approach devoted to meet such tasks is critical for achieving the goals pursued by WBIS. Therefore, in this article an approach oriented to elicit, state, and administrate user knowledge is outlined. This work introduces a user model, which supports the selection of teaching experiences that are delivered to stu- dents in the e-Learning field. The aim is to enhance the apprenticeship of individuals that receive lectures according to the user model that a Web-based Education System (WBES) holds about them. According to a sort of empirical outcomes, it is concluded that: ‘‘The success of WBIS is biased by the accurately acqui- sition, representation, and management of user knowledge fulfilled by the approach”. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction As a result of the Artificial Intelligence (AI) evolution and the spread of Internet, a mutual bias has grown from some of their trends. Actually, several AI scopes are enhanced when their speci- fications include the implementation of AI applications on the Web. Also, new Internet research lines have emerged in order to provide intelligent services, such as semantic Web and wisdom Web (Liu, 2003). While, Internet technologies tend to provide intel- ligent capabilities to a world community of users – which is increasing in number, demands, and services –, the AI research and applications pursue to spread their findings and outcomes world wide. Hence, WBIS have emerged like a kind of ‘‘joint ven- ture” approach that takes into account paradigms from AI and Internet fields. Before a plenty of users with different backgrounds, customs, and preferences, WBIS face up constraints and needs that traditional ‘‘one-size-fits-all” approaches are unable to satisfy (Brusilovsky, 1996). Thereby, WBIS require the support of user knowledge that re- veals who is each user. Such knowledge is set and managed by a user model, which embraces a knowledge repository and an engine. A user model contains a sort of descriptions of what is considered rel- evant about people. It also provides advice to WBIS in order to make easy their adaptation to each individual (Koch, 2000). Thus, WBIS must include a user model with the purpose of customize their pre- sentation, content, navigation, services, and interaction with users. Consequently, WBIS that take into account a user model, behave in an adaptive and intelligent way before people. The cycle process for user modeling embraces three tasks: knowledge acquisition, knowledge representation, and knowledge management. This cycle is accomplished through several itera- tions, from a basic version to a suitable one that is appropriate for being used in real scenarios. Afterwards, the user model is con- tinually updated as a result of future system maintenances. In regards to the knowledge acquisition, this task is devoted to transfer knowledge from one or more sources to repositories man- aged by the engine of the user model. According to Chin (1993), the techniques for acquiring user knowledge are characterized along several orthogonal dimensions as follows: (1) active or passive, depending on the intensity of the user’s participation during the acquisition; (2) automatic or manual, related to the degree of auto- mation of the acquisition task fulfilled by the system; (3) explicit or implicit, according to the type of user feedback; (4) online or offline, based on when the acquisition is performed; (5) adaptive or adapt- able, whether the system or the individual updates the user model respectively. In regards to machine learning for user modeling, Webb, Pazzani, and Billsus (2001) point out four issues to be considered: large data sets, labeled data, concept drift, and 0957-4174/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2009.07.047 * Address: WOLNM, 31 julio 1859, # 1099 B, Leyes Reforma, Iztapalapa, DF 09310, Mexico. Tel.: +52 55 5694 0916/5454 2611; fax: +52 55 5694 0916. E-mail addresses: apenaa@ipn.mx, apenaa@wolnm.org, apenaa@sagitario.cic. ipn.mx. URL: http://www.wolnm.org/apa. Expert Systems with Applications 37 (2010) 2255–2264 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa