ARCADE: A City-Scale Corpus for Fine-Grained Arabic Dialect Tagging
Abstract
The Arabic language is characterized by a rich tapestry of regional dialects that differ substantially in phonetics and lexicon, reflecting the geographic and cultural diversity of its speakers. Despite the availability of many multi-dialect datasets, mapping speech to fine-grained dialect sources, such as cities, remains underexplored. We present ARCADE (Arabic Radio Corpus for Audio Dialect Evaluation), the first Arabic speech dataset designed explicitly with city-level dialect granularity. The corpus comprises Arabic radio speech collected from streaming services across the Arab world. Our data pipeline captures 30-second segments from verified radio streams, encompassing both Modern Standard Arabic (MSA) and diverse dialectal speech. To ensure reliability, each clip was annotated by one to three native Arabic reviewers who assigned rich metadata, including emotion, speech type, dialect category, and a validity flag for dialect identification tasks. The resulting corpus comprises 6,907 annotations and 3,790 unique audio segments spanning 58 cities across 19 countries. These fine-grained annotations enable robust multi-task learning, serving as a benchmark for city-level dialect tagging. We detail the data collection methodology, assess audio quality, and provide a comprehensive analysis of label distributions. The dataset is available on: https://huggingface.co/datasets/riotu-lab/ARCADE-full
The following analysis is generated by AI and may contain errors. It is separate from the authors’ abstract and is not peer review.
Research summary
The main contribution of this paper is the introduction of ARCADE, a novel Arabic speech dataset that enables fine-grained dialect tagging at the city level, addressing a critical gap in existing resources. This work is significant as it not only enhances the understanding of dialectal variation in Arabic but also provides a valuable resource for future research in speech processing and machine learning applications.
Methodology
The methodology employed in the ARCADE project is robust, focusing on the systematic collection of Arabic radio speech from diverse sources. The authors developed a comprehensive data pipeline that ensures high-quality audio segments are captured and annotated by trained native speakers. The annotation process is particularly noteworthy, as it incorporates multiple layers of metadata, including emotional context and dialect categorization, which enhances the dataset's utility for fine-grained dialect tagging. The use of a web-based interface for annotation is a positive aspect, ensuring consistency and efficiency in the labeling process.
Results & evaluation
While the paper does not present experimental results or benchmarks, it provides a thorough analysis of the dataset's composition and the quality of annotations. The authors detail the geographic distribution of recordings and the emotional characteristics of the speech segments, which are critical for future modeling tasks. However, the lack of direct experimental validation of the dataset's effectiveness in dialect identification is a notable gap.
Reproducibility
The authors have made significant efforts to ensure reproducibility by providing transparent documentation of their data collection and annotation protocols. The dataset is openly available on Hugging Face, which facilitates access and encourages further research. However, the paper could benefit from including more implementation details or code to assist researchers in replicating the data collection process.
Limitations
The paper acknowledges several limitations, including geographic and emotional imbalances in the dataset. Some cities are overrepresented, while others lack sufficient recordings. Additionally, the audio quality varies significantly across different regions, which could affect model performance. The authors also note that the dataset is not the largest in terms of duration or number of recordings, which may limit its applicability in certain contexts.
Broader impact
The ARCADE dataset has the potential to significantly advance research in Arabic dialect identification and related fields. By providing fine-grained city-level annotations, it opens up new avenues for sociolinguistic studies and the development of more nuanced speech recognition systems. The dataset's availability can catalyze further research into multi-task learning frameworks and other applications in audio processing. The main contribution of this paper is the introduction of ARCADE, a novel Arabic speech dataset that enables fine-grained dialect tagging at the city level, addressing a critical gap in existing resources. This work is significant as it not only enhances the understanding of dialectal variation in Arabic but also provides a valuable resource for future research in speech processing and machine learning applications.
Author of this paper? Send a correction.