Scholarlink

<p>This paper introduces <i>DarijaMachaair</i>, a large-scale, multi-dimensional dataset for sentiment, emotion, and context analysis in Moroccan Darija, a widely spoken yet underexplored Arabic dialect. The dataset comprises over 117,870 social media comments collected from Facebook, YouTube, Instagram, and TikTok, covering multiple domains. Each instance was manually annotated by trained native speakers across three complementary dimensions: sentiment polarity (positive, negative, neutral), fine-grained emotions (anger, love, joy, neutral, sadness, confusion, optimism, fear, surprise, and ambiguity), and pragmatic contexts (disappointment, appeasing, neutral, sarcastic, and defensive). DarijaMachaair stands out for supporting both Arabic and Arabizi scripts and for capturing authentic emotional expressions grounded in real-world Moroccan cultural discourse. With its combination of large scale and rich multi-dimensional annotations, the dataset provides a valuable benchmark for dialect-specific and emotion-aware natural language processing. This resource fills an important gap in Arabic NLP, particularly for North African dialects, enabling future research on emotionally intelligent applications such as chatbots, moderation tools, and computational sociolinguistic studies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Darijamachaair: a large-scale multi-dimensional emotion and context dataset for Moroccan dialect social media texts

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Sara El Ouahabi,
  • Safaa El Ouahabi,
  • El Wardani Dadi

摘要

This paper introduces DarijaMachaair, a large-scale, multi-dimensional dataset for sentiment, emotion, and context analysis in Moroccan Darija, a widely spoken yet underexplored Arabic dialect. The dataset comprises over 117,870 social media comments collected from Facebook, YouTube, Instagram, and TikTok, covering multiple domains. Each instance was manually annotated by trained native speakers across three complementary dimensions: sentiment polarity (positive, negative, neutral), fine-grained emotions (anger, love, joy, neutral, sadness, confusion, optimism, fear, surprise, and ambiguity), and pragmatic contexts (disappointment, appeasing, neutral, sarcastic, and defensive). DarijaMachaair stands out for supporting both Arabic and Arabizi scripts and for capturing authentic emotional expressions grounded in real-world Moroccan cultural discourse. With its combination of large scale and rich multi-dimensional annotations, the dataset provides a valuable benchmark for dialect-specific and emotion-aware natural language processing. This resource fills an important gap in Arabic NLP, particularly for North African dialects, enabling future research on emotionally intelligent applications such as chatbots, moderation tools, and computational sociolinguistic studies.