A Framework to Develop Intelligent Agents to Support Sentiment Analysis

International audienceMany people including elderly people now live alone with a high tendency toward depression diseases including loneliness. Most of the times social networks are their unique way of expressing their sentiments and speaking freely about their feelings and intentions. However their posts are mostly seen by non-specialists and important information is lost or the adequate reaction is not provided. The aim of this paper is to propose a framework for developing intelligent agents that will be trained to crawl the social networks, to analyse sentiments according with posted expressions which will be classified according with an ontology of emoticons and send alerts to a specialist who will provide the adequate psychologic and social support. These intelligent agents can work like a team, developing collaboration work and developing skills and competences similar to Darwin’s theory of biological evolution. The expressions’ classification will be a priori annotated by a panel of experts and these annotation will be used to train the intelligent agents. The system will also use an ontology not only to reason the concepts but also to find new association of terms and expressions that will be presented to the experts to get their classifications. The final outcomes will be an evolutive eco system of intelligent agents capable of gathering knowledge and using their competences to analyse sentiments according to an ontology of emotions The proposed framework is based on a literature review, technological state of the art and includes the global architecture, technology and implementation training to get a fully functional system. Some foundation concepts, methods and technologies will be reused in this framework including emotion ontology, and ontology editor Protégé, a text mining information extractor and classifier and a platform to develop intelligent agents, JADE