Identificação de características de aspectos emocionais associados a elementos de narrativas audiovisuais

Abstract

Social networks such as YouTube have increased the ability to disseminate content in different formats. Due to the narrative character that these contents have, automatically

identifying and extracting elements from the video can be a solution for the charac- terization of narratives more effectively. This work presents the proposal to identify

characteristics of emotional aspects in audiovisual narratives, through the manipulation of computational libraries and pre-trained machine learning models for the creation of a Dataframe, containing the emotion information identified through detection and facial recognition and recognition of facial expressions of characters found in videos. From the results obtained, as a way of validating the proposal, a comparison was made of the characteristics extracted manually by a professional in the communication area, with the characteristics identified and extracted by the proposed architecture, which showed the possibility of optimizing the process of extracting characteristics of YouTube videos for the analysis of audiovisual narratives, through the automation of part of this process.