{"slug": "learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to", "title": "Learning the Arabic Dialect Continuum as a Continuous Space: A Regression Approach to Speaker Origin Prediction", "summary": "A regression-based approach to Arabic dialect geolocation that models dialectal variation as a continuous geographic space achieves a median localization error of 481.2 km under a leakage-free 5-fold GroupKFold protocol, according to a new arXiv preprint (2607.19751v1). The hierarchical neural architecture fuses XLS-R-300M and Whisper-large-v3 encoder representations with phonotactic descriptors, and a spherical geodesic loss optimizes great-circle distance. Under a zero-shot city-masking protocol, mean error rises to 1173.3 km, a 1.32x degradation relative to seen cities.", "body_md": "arXiv:2607.19751v1 Announce Type: new\nAbstract: We present a regression-based approach to Arabic dialect geolocation that models dialectal variation as a continuous geographic space rather than discrete categories. Speaker origin is predicted as continuous latitude-longitude coordinates using a hierarchical neural architecture that fuses frame-level XLS-R-300M and Whisper-large-v3 encoder representations with phonotactic descriptors through a Transformer encoder and a learnable attention-pooled query. A spherical geodesic loss directly optimizes great-circle distance on Earth's surface, avoiding distortions inherent to planar coordinate regression. Under a leakage-free 5-fold GroupKFold protocol grouped by source recording, our model attains a pooled median localization error of 481.2 km. Auxiliary country and city heads reach 64.5% and 45.2% accuracy, respectively. A permutation Mantel test on the learned latent space provides quantitative support for the Arabic dialect continuum hypothesis. To probe true generalization, we further introduce a city-masking protocol in which two cities per fold are removed from training but retained in validation. Under this zero-shot regime, the mean error rises to 1173.3 km, a 1.32x degradation relative to seen cities. Our findings establish continuous geographic modeling as a principled framework for Arabic dialect geolocation and quantify both its strengths and the substantial headroom that remains.", "url": "https://wpnews.pro/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to", "canonical_source": "https://www.machinebrief.com/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-iadr", "published_at": "2026-07-23 04:00:00+00:00", "updated_at": "2026-07-23 04:04:55.279767+00:00", "lang": "en", "topics": ["machine-learning", "natural-language-processing", "artificial-intelligence"], "entities": ["arXiv", "XLS-R-300M", "Whisper-large-v3"], "alternates": {"html": "https://wpnews.pro/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to", "markdown": "https://wpnews.pro/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to.md", "text": "https://wpnews.pro/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to.txt", "jsonld": "https://wpnews.pro/news/learning-the-arabic-dialect-continuum-as-a-continuous-space-a-regression-to.jsonld"}}