Towards personality classification through arabic handwriting analysis

Abstract

Classification of personality types based on Arabic handwriting is a challenging task. It has multiple practical application since handwriting is a unique attribute for each person that provides as much differentiation as fingerprints. Accurate analysis of handwritten documents is useful in diverse areas including human resource management, criminal justice, forensics, security, archaeology, and countless other spheres of life. Writer identification is a big challenge mostly due to a limited human capability for observing and recognizing different styles of writing. Arabic is a language used by millions of people. Recognition of handwritten Arabic characters would, therefore, be of tremendous help in various sectors like mail sorting, verification of checks, etc. even in countries where Arabic is used only occasionally. For this purpose, we collected around 160 samples of Arabic handwriting from 83 different writers, where every writer wrote the same paragraph in both normal and fast writing. Carl Jung’s and Isabel Briggs Myers’ personality type theory with 16 different personality type was used for classifying the writers. The writers were asked to complete the personality test that contains 64 questions related to Jung-Myers’ typology. Specific features of Arabic language handwriting were used to match each written document with a personality type of the writer based on a set of classification techniques. We used four different machine learning classification algorithms with 4-fold cross validation technique. We conducted two experiments and achieved a maximum accuracy of 68.67%.