Hypertext Classification: A Multiagent Model for Adaptive Learning Algorithm

The problem of hypertext classification deals with objects possessing more complex information structure than the plain text has. Present hypertext classification systems show their effectiveness on invariant data retrieval but in circumstances of topic modification they are not ready for classifier adjustment. To achieve a higher level of personalization in work with automated classifier a system with a feedback is proposed. A model is described that complements the traditional Naive Bayes classification algorithm with a dynamically retainable scheme working with various features of hypertext data and uses a multiagent approach. A prototype of a system contains the three agent types: user interface agent, classification agent, and mediation agent. Implementation of a system provides a user or user group with a possibility of dynamic adjustment of the structure of a classifier via activity of multiagent system. KEYWORDS Multiagent system, hypertext classification, Web, dynamic classif...

• Type: Conference
• Year: 2003
• Where: IADIS
• Authors: Stanislav Baglei
• Keywords: Classifier Adjustment, Hypertext Classification, Hypertext Classification Deals, IADIS 2003, IADIS 2008