Muhammad Abdul-Mageed · page 3
Papers
- Benchmarking LLaMA-3 on Arabic Language Generation Tasks (2024 · 12 citations)
This study asks how well the strongest open-weight model of its time, LLaMA-3-70B, generates Arabic text, a question that earlier evaluations had left open. - Towards Zero-Shot Text-To-Speech for Arabic Dialects (2024 · 10 citations)
Arabic lacks open research on zero-shot multi-speaker text-to-speech (ZS-TTS), where a system must imitate a voice it never saw in training from only a few seconds of reference audio. - Arabic Automatic Story Generation with Large Language Models (2024 · 9 citations)
Arabic has lagged behind English in LLM-based story generation, mostly because short Arabic story data is scarce. - AraHealthQA 2025: The First Shared Task on Arabic Health Question Answering (2025 · 8 citations)
AraHealthQA 2025 is a shared task held with ArabicNLP 2025 at EMNLP 2025. It addresses the shortage of Arabic medical question-answering resources and standardized ways to compare systems. It has two tracks: - To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation (2024 · 8 citations)
Multilingual speech recognisers such as Whisper and SeamlessM4T are large, costly to run, and mostly tested on Modern Standard Arabic. - BERT-Based Arabic Social Media Author Profiling (2019 · 8 citations)
The paper describes a system for the APDA shared task on Arabic author profiling. Given 100 tweets from each user, it predicts three attributes: age group (3 classes), country-level dialect (15 classes), and gender (binary). The authors fine-tune multilingual BERT separately for each attribute. Classification happens… - A Benchmark Study of Contrastive Learning for Arabic Social Meaning (2022 · 7 citations)
The paper asks whether supervised contrastive learning (CL), which has helped many NLP tasks, also helps Arabic social meaning tasks such as dialect identification, emotion recognition, abusive and hate speech detection, and demographic attribute prediction. - Arab Voices: Mapping Standard and Dialectal Arabic Speech Technology (2026 · 3 citations)
Speech resources for Dialectal Arabic (DA) are scattered and inconsistent. Corpora use different dialect labeling schemes (country names, broad regions, ISO codes), cover different domains, and are recorded under different conditions. This makes it hard to compare datasets or evaluate models fairly. The authors… - From Nile Sands to Digital Hands: Machine Translation of Coptic Texts (2024 · 3 citations)
Coptic, the late stage of the Egyptian language, is still used in church liturgy, but it has almost no digital resources and current language technology largely ignores it. - On the Utility of Pretraining Language Models on Synthetic Data (2024 · 3 citations)
Pretraining language models normally needs very large amounts of clean text, which low-resource languages often lack. The authors test whether Arabic models can be pretrained mostly or entirely on machine-produced text. They build four synthetic corpora: - DiaLex: A Benchmark for Evaluating Multidialectal Arabic Word Embeddings (2021 · 3 citations)
Arabic dialects had no shared way to check whether word embeddings capture their vocabulary and grammar. Earlier analogy benchmarks for Arabic covered only Modern Standard Arabic (MSA) or were translated from English. DiaLex fills this gap. It is a hand-built set of word pairs for Algerian, Egyptian, Lebanese, Syrian… - AlexandriaX 2026: The First Shared Task on Dialectal Arabic Machine Translation (2026 · 2 citations)
AlexandriaX 2026 is a shared task on machine translation for Arabic dialects. Most Arabic tools were built for MSA. Dialect output can keep the meaning and still sound wrong, for example through incorrect gender agreement, the wrong level of formality, or MSA slipping in. Aggregate metrics also say little about why… - Speech Technology for Everyone: Automatic Speech Recognition for Non-Native English (2021 · 2 citations)
English speech recognizers work noticeably worse for non-native (L2) speakers, and there is little annotated accented speech to fix this. - Bulbul: A Dataset for Dialectal Arabic Speech Recognition (2026 · 1 citation)
Arabic speech recognition struggles because speech resources are scarce and skewed toward a few varieties: MSA, Egyptian, broad Levantine, Gulf and Moroccan. - Halluscoring 2026: The first shared task on llms hallucination detection and answer verification (2026)
HalluScoring 2026 is a shared task on whether systems can tell when an LLM's answer to an Arabic question contains fabricated or wrong facts, and whether they can then recover the true answer. - Zero-Shot Context-Aware ASR for Diverse Arabic Varieties (2026)
Multilingual speech recognizers such as Whisper handle Modern Standard Arabic (MSA) reasonably well, but their accuracy drops sharply on regional dialects and accented speech, where labeled data is scarce.
This page in Markdown → https://md.research.elamur.ai/free/en/authors/muhammad-abdul-mageed/page/3/