Learning the Arabic Dialect Continuum as a Continuous Space: A Regression Approach to Speaker Origin Prediction

2026-07-22

AI Summary

This study addresses the limitations of traditional Arabic dialect geolocation approaches, which treat dialects as discrete categories and fail to capture their inherent continuous geographic variation. To overcome this, the authors propose the first end-to-end continuous geolocation regression framework that models dialects as a geographic continuum. The model integrates speech representations from XLS-R-300M and Whisper-large-v3 with phonological features, employing a Transformer architecture and learnable attention pooling to directly predict speaker coordinates. A spherical geodesic loss is introduced to optimize great-circle distance. Evaluated under a strict no-data-leakage 5-fold GroupKFold protocol, the approach achieves a median localization error of 481.2 km, with country- and city-level classification accuracies of 64.5% and 45.2%, respectively. In zero-shot city-masking tests, the error increases to 1173.3 km.

Abstract

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.

Problem

• Arabic dialect continuum
• geolocation
• continuous space
• speaker origin prediction
• dialectal variation

Innovation

• continuous dialect modeling
• spherical geodesic loss
• hierarchical neural fusion
• zero-shot city masking
• Arabic dialect continuum