Personality Analysis from Lip Shapes: Marifetnâme and Explainable Artificial Intelligence Approach

Abstract

Personality analysis provides valuable information about the inner world of an individual by revealing unknown aspects of their psychological structure, behaviors, and character traits. Throughout history, various disciplines have developed different methods to determine personality traits. In traditional methods, personality analysis has usually been conducted using techniques such as surveys, interviews, and observation. However, since these methods are mostly based on subjective evaluations, they have limitations in terms of accuracy and reliability of the results. Algorithmbased studies developed to overcome these limitations focus on temporary facial features and demographic features that can be easily extracted from faces rather than morphological features. In this study, mouth and lip descriptions in Ibrahim Hakk's work Marifetname were examined with modern artificial intelligence technologies. Face images taken from VGGFace2 dataset were processed using Haar Cascade Classifier and facial landmarks system, then trained with three different CNN architectures (AlexNet, ResNet, YOLO). YOLO model was preferred due to its high detection success. The results were analyzed with explainable artificial intelligence techniques and the reliability of the model was confirmed. The study provides a new perspective to traditional personality analysis using artificial intelligence technologies and makes significant contributions to this field. © 2024 IEEE.