RESILIENCE UNDER FIRE: DEVELOPING AN ANNOTATED DATASET OF HOPE SPEECH IN ARABIC YOUTUBE COMMENTS DURING THE GAZA GENOCIDE

Abstract

Social media has become a critical space for discourse during conflicts, shaping narratives, influencing public sentiment, and providing a platform for alternative perspectives. Since the beginning of the Protel war on Gaza that started in October 2023, more than 48000 Palestinians have been martyred and over 2 million displaced from their homes due to Israeli occupation forces air strikes. During this time, social media platforms have been used as a medium for people around the world to share videos of crimes perpetrated in Gaza and to show their support for its people. This study explores specifically hope speech in YouTube comments related to the Gaza Genocide during the period (2023-2024), analyzing how users engage in expressions of resilience, solidarity, and optimism amid crises. Using a dataset of 10,000 Arabic-language comments, this research employs a detailed annotation framework to classify discourse into hope speech, no hope speech, and neutral/unclear speech categories. The findings reveal that hope speech dominated (64.3%), primarily in the form of religious and spiritual encouragement, while no hope speech (13.5%) was characterized by frustration, disillusionment with leadership, and criticism without solutions. The study also highlights annotation challenges, particularly in dealing with Arabic dialects, sarcasm, and implicit sentiment, underscoring the limitations of current NLP models in accurately detecting nuanced emotions in Arabic social media discourse. These results align with prior research on digital conflict narratives, including Liyih et al. (2024), which found that social media users present a more complex, emotionally charged perspective than mainstream media’s binary portrayals of war. This study contributes to the growing field of hope speech detection and emphasizes the need for advanced NLP models capable of handling Arabic dialects in sentiment analysis.

Date of Award	2025
Original language	American English
Awarding Institution	HBKU College of Humanities and Social Science