Modelo para identificação de contexto social através da inferência de interações sociais

Abstract

Context-aware Computing is characterized by the development of devices capable of making decisions, offering services and answers based on the user’s current context. It is formed by different types of context, among them there is the social context. The main factor of a social context is information about social relations. These relationships are formed by a set of social interactions. Therefore, in this work we suggest the SocialCount, a model of inference of social interactions face to face for the identification of a social context. The challenge of the area is the use of abstract concepts (social interaction, social relation, social context) in computational means. For the inference of social interactions we use a set of approaches, the differential in relation to related works is the speaker recognition. The identification of the social context is performed based on three factors: number of users in the group, interactions between users and the main type of relationship present in the group. In the experiment performed, the set of approaches used by SocialCount inferred enough interactions to achieve an accuracy of 86% in the classification of relationships. With respect to the identification of the social context, the context identified by SocialCount obtained two factors (number of users in the group and interactions between users) equivalent to the context recognized by the validator method. Thus, the model was able to achieve the proposed goal of adequately inferring face-to-face social interactions made by the user, identifying their social context.