Scholarlink

This paper proposes an ontology-based method for extracting semantic triples from Arabic news articles, addressing unique challenges of Arabic text processing. Our approach combines named entity recognition, part-of-speech tagging, dependency parsing, and sentiment analysis, mapped to a domain ontology designed for conflict analytics. Our methodology enhances Arabic NLP tools with domain-specific rules, implementing a hybrid architecture integrating CAMeL Tools and FARASA for improved entity extraction. Experimental results on 5,000 Modern Standard Arabic news articles demonstrate 96.06% named entity coverage, with balanced recognition across locations (43.5%), organizations (25.7%), and persons (24.8%). Network analysis reveals hierarchical information patterns, with news articles serving as central nodes and locations as contextual anchors, forming distinct communities of entities. Furthermore, sentiment analysis shows predominant neutral (51.3%) and positive (45.3%) polarities despite the conflict context, reflecting journalistic objectivity. Despite current limitations in Arabic NLP tools and resources, our approach demonstrates effective semantic knowledge extraction through combined ontological and sentiment frameworks

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

SPARKLE: Structured Parsing for Arabic Resource Knowledge and Language Extraction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Fouad Al Tfaily,
  • Hussein Hazimeh,
  • Hassan Harb,
  • Karl Daher,
  • Omar Abou Khaled,
  • Elena Mugellini,
  • Ali Jaber,
  • Ali Takach

摘要

This paper proposes an ontology-based method for extracting semantic triples from Arabic news articles, addressing unique challenges of Arabic text processing. Our approach combines named entity recognition, part-of-speech tagging, dependency parsing, and sentiment analysis, mapped to a domain ontology designed for conflict analytics. Our methodology enhances Arabic NLP tools with domain-specific rules, implementing a hybrid architecture integrating CAMeL Tools and FARASA for improved entity extraction. Experimental results on 5,000 Modern Standard Arabic news articles demonstrate 96.06% named entity coverage, with balanced recognition across locations (43.5%), organizations (25.7%), and persons (24.8%). Network analysis reveals hierarchical information patterns, with news articles serving as central nodes and locations as contextual anchors, forming distinct communities of entities. Furthermore, sentiment analysis shows predominant neutral (51.3%) and positive (45.3%) polarities despite the conflict context, reflecting journalistic objectivity. Despite current limitations in Arabic NLP tools and resources, our approach demonstrates effective semantic knowledge extraction through combined ontological and sentiment frameworks