Browsing by Author "Dehkharghani, R."

Now showing 1 - 2 of 2

Exploring Isis’ Takfir Discourse: A BERT-Based Entity Level Sentiment Analysis Approach
(Springer, 2025) Dehkharghani, R.; Aydin, M.N.; Yıldırım, Ş.
The Islamic State of Iraq and Syria (ISIS) significantly influenced the lives of many people during and after the Syrian civil conflict, especially civilians. Analyzing social media discussions about ISIS can provide valuable insights into the group’s beliefs and attitudes. In this paper, we examine ISIS’s takfir discourse—their practice of labeling other Muslims as unbelievers to justify exclusion or violence—in Telegram groups. We collected 14,500 Telegram messages (2015–2017) using snowball sampling, API-based crawling, language filtering, and time-window selection. We then integrated a BERT-based Named Entity Recognition (NER) model with two layers of the Span ASTE (Aspect-Based Sentiment Analysis) model. We also used the Span ASTE as an end-to-end baseline for comparison. Based on Precision, Recall, and F1-scores, our hybrid model outperformed the baselines, demonstrating its effectiveness in sentiment analysis of extracted named entities. © 2025 Elsevier B.V., All rights reserved.
Multi-Task Learning on Mental Disorder Detection, Sentiment Analysis, and Emotion Detection Using Social Media Posts
(Institute of Electrical and Electronics Engineers Inc., 2024) Armah, C.; Dehkharghani, R.; Computer Engineering; 05. Faculty of Engineering and Natural Sciences; 01. Kadir Has University
Mental disorders such as suicidal behavior, bipolar disorder, depressive disorders, and anxiety have been diagnosed among the youth recently. Social media platforms such as Reddit have become popular for anonymous posts. People are far more likely to share on these social media platforms what they really feel like in their real lives when they are anonymous. It is thus helpful to extract people's sentiments and feelings from these platforms in training models for mental disorder detection. This study uses multi-task learning techniques to examine the estimation of behaviors and mental states for early mental disease diagnosis. We propose a multi-task system trained on three related tasks: mental disorder detection as the primary task, emotion analysis, and sentiment analysis as auxiliary tasks. We took the SWMH dataset, which included four main different mental disorders already labeled (bipolar, depression, anxiety, and suicide) and offmychest. We then added labels for emotion and sentiment to the dataset. The observed results are comparable to previous studies in the field and demonstrate that deep learning multi-task frameworks can improve the accuracy of related text classification tasks when compared to training them separately as single-task systems. © 2024 IEEE.