Skip to main content
Thesis defences

PhD Oral Exam - Afnan Garoot, Computer Science and Software Engineering

Handwriting Analysis and Personality: A Computerized Study on Validity of Graphology


Date & time
Friday, October 1, 2021 (all day)
Cost

This event is free

Organization

School of Graduate Studies

Contact

Dolly Grewal

Where

Online

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

Handwriting analysis, also known as graphology, is defined as an analysis of a psychological structure of a human subject through his or her handwriting. Graphologists use handwriting as an indication of personality traits represented by neurological patterns in the brain such as emotional outlay, fears, honesty and defences but not identifying the writer’s age, race, religion, or nationality. It is performed manually in two steps, firstly, extracting specific features from a handwriting sample such as the size of writing, pressure, and spaces between words or letters. Then, interpreting the extracted features based on graphological rules.

Handwriting analysis has been applied recently in different fields, especially where making a crucial decision is highly desirable. The most common fields are forensic evidence, criminology, disease analysis, therapy, and personality prediction. However, making a crucial decision based on the results of handwriting analysis is a controversial scientific topic because the validity of graphology rules is still in question. Therefore, there is a need to conduct more studies on examining the validity of handwriting analysis by researchers.

A few validity studies on handwriting analysis have already been conducted earlier using the evaluation of correlation between psychological questionnaires and manual handwriting analysis and they ended up with conflicting results. Manual graphology is a tedious, subjective, and error prone task. Sometimes it leads to inconsistent predictions and conclusions between graphologists. So, using manual handwriting analysis for evaluating the validity of graphology could be the reason of the conflicting conclusions obtained by the early studies. Therefore, there is a need for a reliable investigation method in order to validate graphology more accurately. For this, replacing manual handwriting analysis system with a computerized one in validation studies would be the ideal solution.

In this research, we conduct an empirical study that investigates the validation of graphology rules. For this, the study evaluates the correlation between one of psychological tests named Big Five Factor Markers Test (BFFMT) and our proposed automated handwriting analysis system that measures the level of the same big five personality traits based on graphological rules. The big five factor model is a set of five broad trait dimensions which are Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Open to Experience.

In this work, we propose an automatic graphology system that implements multi-label algorithm on an ensemble learning model named AvgMlSC. AvgMlSC deploys two learning-based classifiers which are Multi-label Support Vector Machine (MLSVM) and Multi-label Convolutional Neural Network (MLCNN) based on off-line handwriting recognition. The averaging ensemble method was employed along with Multi-label Synthetic Minority Over-sampling Technique (MLSMOTE) in order to handle the issue of imbalanced dataset. The results of our experiments validate the potential of our proposed model in comparison to five baseline classifiers i.e., Logistic Regression (LR), Naïve Bayes (NB), K-Neighbors (KN), Support Vector Machine (SVM), and Convolutional Neural Network (CNN). Moreover, the results are compared to two studies from the state-of-the-art that have used the same form of data as in our model and the same performance measures to evaluate their proposed models.

The proposed system is trained and tested using our own dataset named Handwriting for the Big Five Factors (HWBFF) collected by a survey. It consists of 1066 handwriting samples written in English, French, Chinese, Arabic, and Spanish. The results reveal that the proposed ensemble learning model outperformed the overall performance of the five traditional models with 93% predictive accuracy, 0.94 Area Under the Receiver Operating Characteristic Curve (AUC), and 90% F-Score. In addition, it achieved higher values than the state-of-the-art with a remarkable increase in accuracy (84%, 92%, 96%, 99%, and 96%) and F-Score (68%, 90%, 97%, 99%, and 97%) for Extraversion, Conscientiousness, Emotional Stability, Agreeableness and Open to Experience, respectively.

After getting a highest performance evaluation for AvgMlSC, the predicted results of the big five factors generated by the proposed automated graphology system were correlated with the scores of the BFFMT using Spearman’s rho (ρ) correlation coefficients in order to evaluate the validity of handwriting analysis. The statistical test reports there is a statistically significant relationship between the score of Big Five Factors questionnaire and the graphologist’s evaluation for the Big Five Factors with a different strength of relationship as follows. A weak positive relationship is found for Extraversion. However, a moderate positive relationship is reported for Conscientiousness and Open to Experience. On the other hand, a strong positive relationship is indicated for Agreeableness. Whilst a very weak positive relationship has been found for the last factor which is Emotional Stability.

In this study, the BFF correlation of printed and cursive writing for Latin languages are compared. The results show that the strength of the BFF correlation for printed writing is lower than cursive writing for the five factors. This difference is expected because printed writings are often artificial and are often associated with a persona personality.

Back to top

© Concordia University