TY - GEN
T1 - Depression Detection in Social Media Using a Psychoanalytical Technique for Feature Extraction and a Cognitive Based Classifier
AU - Hosseini-Saravani, Seyed Habib
AU - Besharati, Sara
AU - Calvo, Hiram
AU - Gelbukh, Alexander
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020
Y1 - 2020
N2 - Depression detection in social media is a multidisciplinary area where psychological and psychoanalytical findings can help machine learning and natural language processing techniques to detect symptoms of depression in the users of social media. In this research, using an inventory that has made systematic observations and records of the characteristic attitudes and symptoms of depressed patients, we develop a bipolar feature vector that contains features from both depressed and non-depressed classes. The inventory we use for feature extraction is composed of 21 categories of symptoms and attitudes, which are primarily clinically derived in the course of the psychoanalytic psychotherapy of depressed patients, and systematic observations and records of their characteristic attitudes and symptoms. Also, getting insight from a cognitive idea, we develop a classifier based on multinomial Naïve Bayes training algorithm with some modification. The model we develop in this research is successful in classifying the users of social media into depressed and non-depressed groups, achieving the F1 score 82.75%.
AB - Depression detection in social media is a multidisciplinary area where psychological and psychoanalytical findings can help machine learning and natural language processing techniques to detect symptoms of depression in the users of social media. In this research, using an inventory that has made systematic observations and records of the characteristic attitudes and symptoms of depressed patients, we develop a bipolar feature vector that contains features from both depressed and non-depressed classes. The inventory we use for feature extraction is composed of 21 categories of symptoms and attitudes, which are primarily clinically derived in the course of the psychoanalytic psychotherapy of depressed patients, and systematic observations and records of their characteristic attitudes and symptoms. Also, getting insight from a cognitive idea, we develop a classifier based on multinomial Naïve Bayes training algorithm with some modification. The model we develop in this research is successful in classifying the users of social media into depressed and non-depressed groups, achieving the F1 score 82.75%.
KW - Depression detection
KW - Natural Language Processing
KW - Naïve Bayes
KW - Psychoanalysis
KW - Rational Speech Act
KW - Social media
UR - http://www.scopus.com/inward/record.url?scp=85092891445&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-60887-3_25
DO - 10.1007/978-3-030-60887-3_25
M3 - Contribución a la conferencia
AN - SCOPUS:85092891445
SN - 9783030608866
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 282
EP - 292
BT - Advances in Computational Intelligence - 19th Mexican International Conference on Artificial Intelligence, MICAI 2020, Proceedings
A2 - Martínez-Villaseñor, Lourdes
A2 - Ponce, Hiram
A2 - Herrera-Alcántara, Oscar
A2 - Castro-Espinoza, Félix A.
PB - Springer Science and Business Media Deutschland GmbH
T2 - 19th Mexican International Conference on Artificial Intelligence, MICAI 2020
Y2 - 12 October 2020 through 17 October 2020
ER -