Convex Low-resource Accent-Robust Language Detection in Speech Recognition

Globalization and multiculturalism continue to produce increasingly diverse speech varieties. Yet current spoken dialogue systems frequently fail on under-represented dialects and accents, often misidentifying the input language and causing cascading failures in downstream dialogue tasks. Addressing this dialectal variance under low-resource constraints remains an open challenge, as standard fine-tuning is computationally expensive and prone to overfitting on high-dimensional speech data. We propose Convex Language Detection (CLD), a novel framework that integrates theoretically grounded convex optimization techniques into the spoken dialogue systems pipeline. Our method is efficiently implemented via multi-GPU Alternating Direction Method of Multipliers (ADMM) in JAX, thus providing global optimality guarantees and fast training in polynomial time. Theoretically, we prove that our convex objective induces certified margin stability and provide guarantees against feature perturbations. Empirically, we demonstrate sample efficiency and robustness to input dialectical variation, achieving 97–98\% accuracy in challenging low-resource regimes. Our open-source package is available at https://pypi.org/project/jaxcld/.

Modern voice assistants such as Siri and Amazon Echo often struggle to understand people who speak with regional accents or under-represented dialects. When a system misidentifies the language or dialect being spoken, it causes a chain reaction of errors that prevents the model from responding correctly. Fixing this usually requires retraining massive neural networks, which is incredibly expensive and often fails when there is very little data available for a specific accents. To solve this, we introduce a framework called Convex Language Detection (CLD). Instead of relying on traditional training techniques and large data, CLD uses a mathematical derivation to efficiently and accurately route accented input speech. Built using the high-performance JAX framework, our method guarantees finding the absolute best settings to recognize a dialect, even when training data is extremely limited (less than 100 short samples). Our experiments prove that CLD makes voice systems significantly more robust against subtle variations in speech, ensuring that technology works reliably for everyone in a global ecosystem.