Analyzing Web Chat Messages for Recommending Items From a Digital Library

Abstract

This work presents a recommender system that analyzes textual messages sent during a communication session in a private Web chat, identifies the context of each message and recommends items from a Digital Library. Recommendations are directly made to users in the chat screen and are decided by a software system through a proactive paradigm, without any request of the users. A domain ontology, containing concepts and a controlled vocabulary, is used to identify subjects in textual messages and to automatically classify items of the Digital Library.