Academic

ADAB: Arabic Dataset for Automated Politeness Benchmarking -- A Large-Scale Resource for Computational Sociopragmatics

arXiv:2602.13870v1 Announce Type: new Abstract: The growing importance of culturally-aware natural language processing systems has led to an increasing demand for resources that capture sociopragmatic phenomena across diverse languages. Nevertheless, Arabic-language resources for politeness detection remain under-explored, despite the rich and complex politeness expressions embedded in Arabic communication. In this paper, we introduce ADAB (Arabic Politeness Dataset), a new annotated Arabic dataset collected from four online platforms, including social media, e-commerce, and customer service domains, covering Modern Standard Arabic and multiple dialects (Gulf, Egyptian, Levantine, and Maghrebi). The dataset was annotated based on Arabic linguistic traditions and pragmatic theory, resulting in three classes: polite, impolite, and neutral. It contains 10,000 samples with linguistic feature annotations across 16 politeness categories and achieves substantial inter-annotator agreement (ka

H
Hend Al-Khalifa, Nadia Ghezaiel, Maria Bounnit, Hend Hamed Alhazmi, Noof Abdullah Alfear, Reem Fahad Alqifari, Ameera Masoud Almasoud, Sharefah Ahmed Al-Ghamdi · · 1 min read · 10 views
#cs.CL

arXiv:2602.13870v1 Announce Type: new Abstract: The growing importance of culturally-aware natural language processing systems has led to an increasing demand for resources that capture sociopragmatic phenomena across diverse languages. Nevertheless, Arabic-language resources for politeness detection remain under-explored, despite the rich and complex politeness expressions embedded in Arabic communication. In this paper, we introduce ADAB (Arabic Politeness Dataset), a new annotated Arabic dataset collected from four online platforms, including social media, e-commerce, and customer service domains, covering Modern Standard Arabic and multiple dialects (Gulf, Egyptian, Levantine, and Maghrebi). The dataset was annotated based on Arabic linguistic traditions and pragmatic theory, resulting in three classes: polite, impolite, and neutral. It contains 10,000 samples with linguistic feature annotations across 16 politeness categories and achieves substantial inter-annotator agreement (kappa = 0.703). We benchmark 40 model configurations, including traditional machine learning, transformer-based models, and large language models. The dataset aims to support research on politeness-aware Arabic NLP.

Executive Summary

The article introduces ADAB, a comprehensive Arabic dataset designed for automated politeness benchmarking in natural language processing (NLP). Collected from diverse online platforms, ADAB includes 10,000 samples in Modern Standard Arabic and multiple dialects, annotated across three classes: polite, impolite, and neutral. The dataset achieves substantial inter-annotator agreement and benchmarks 40 model configurations, including traditional machine learning and transformer-based models. ADAB aims to advance research in politeness-aware Arabic NLP, addressing the under-explored area of sociopragmatic phenomena in Arabic communication.

Key Points

  • ▸ ADAB is a new Arabic dataset for politeness detection, covering multiple dialects and domains.
  • ▸ The dataset includes 10,000 samples annotated across 16 politeness categories with substantial inter-annotator agreement.
  • ▸ Benchmarking of 40 model configurations demonstrates the dataset's utility for advancing Arabic NLP research.

Merits

Comprehensive Coverage

ADAB covers a wide range of Arabic dialects and domains, making it a valuable resource for understanding politeness in diverse contexts.

High Inter-Annotator Agreement

The dataset achieves a kappa score of 0.703, indicating reliable and consistent annotations.

Benchmarking Multiple Models

The article benchmarks 40 model configurations, providing a robust foundation for future research in politeness detection.

Demerits

Limited Sample Size

While 10,000 samples are substantial, the dataset could benefit from a larger and more diverse sample size to capture the full spectrum of Arabic politeness expressions.

Potential Bias in Data Collection

The data is collected from online platforms, which may introduce biases and limit the generalizability of the findings.

Annotation Complexity

Annotating politeness expressions is inherently subjective and may vary across annotators, despite the high inter-annotator agreement.

Expert Commentary

The introduction of ADAB marks a significant step forward in the field of Arabic NLP, particularly in the under-explored area of politeness detection. The dataset's comprehensive coverage of multiple dialects and domains, along with its high inter-annotator agreement, provides a solid foundation for future research. However, the limited sample size and potential biases in data collection are areas that could be addressed in future iterations. The benchmarking of 40 model configurations demonstrates the dataset's utility and sets a high standard for future work. Overall, ADAB is a valuable resource that will advance the development of culturally-aware NLP systems and contribute to the broader goal of building more nuanced and context-aware language models.

Recommendations

  • ✓ Future research should aim to expand the dataset to include a larger and more diverse sample size to capture a broader range of Arabic politeness expressions.
  • ✓ Efforts should be made to mitigate potential biases in data collection, ensuring that the dataset is representative of various contexts and user groups.

Sources

Related Articles