This is an internal, incomplete preview of a proposed change to the ACL Anthology.
For efficiency reasons, we don't generate MODS or Endnote formats, and the preview may be incomplete in other ways, or contain mistakes.
Do not treat this content as an official publication.
Waad BenKheder
Also published as:
Waad Ben Kheder
Fixing paper assignments
Please select all papers that belong to the same person.
Indicate below which author they should be assigned to.
We present our IWSLT 2025 submission for the low-resource track on North Levantine Arabic to English speech translation, building on our IWSLT 2024 efforts. We retain last year’s cascade ASR architecture that combines a TDNN-F model and a Zipformer for the ASR step. We upgrade the Zipformer to the Zipformer-Large variant (253 M parameters vs. 66 M) to capture richer acoustic representations. For the MT part, to further alleviate data sparsity, we created a crowd-sourced parallel corpus covering five major Arabic dialects (Tunisian, Levantine, Moroccan, Algerian, Egyptian) curated via rigorous qualification and filtering. We show that using crowd-sourced data is feasible in low-resource scenarios as we observe improved automatic evaluation metrics across all dialects. We also experimented with the dataset under a high-resource scenario, where we had access to a large, high-quality Levantine Arabic corpus from LDC. In this setting, adding the crowd-sourced data does not improve the scores on the official validation set anymore. Our final submission scores 20.0 BLEU on the official test set.
This paper presents ALADAN’s approach to the IWSLT 2024 Dialectal and Low-resource shared task, focusing on Levantine Arabic (apc) and Tunisian Arabic (aeb) to English speech translation (ST). Addressing challenges such as the lack of standardized orthography and limited training data, we propose a solution for data normalization in Dialectal Arabic, employing a modified Levenshtein distance and Word2vec models to find orthographic variants of the same word. Our system consists of a cascade ST system integrating two ASR systems (TDNN-F and Zipformer) and two NMT modules derived from pre-trained models (NLLB-200 1.3B distilled model and CohereAI’s Command-R). Additionally, we explore the integration of unsupervised textual and audio data, highlighting the importance of multi-dialectal datasets for both ASR and NMT tasks. Our system achieves BLEU score of 31.5 for Levantine Arabic on the official validation set.
Les applications de compréhension du langage parlé sont moins performantes si les documents transcrits automatiquement contiennent un taux d’erreur-mot élevé. Des solutions récentes proposent de projeter ces transcriptions dans un espace de thèmes, comme par exemple l’allocation latente de Dirichlet (LDA), la LDA supervisée ainsi que le modèle author-topic (AT). Une représentation compacte originale, appelée c-vector, a été récemment introduite afin de surmonter la difficulté liée au choix de la taille de ces espaces thématiques. Cette représentation améliore la robustesse aux erreurs de transcription, en compactant les différentes représentations LDA d’un document parlé dans un espace réduit. Le défaut majeur de cette méthode est le nombre élevé de sous-tâches nécessaires à la construction de l’espace c-vector. Cet article propose de corriger ce défaut en utilisant un cadre original fondé sur un espace de caractéristiques robustes de faible dimension provenant d’un ensemble de modèles AT considérant à la fois le contenu du dialogue parlé (les mots) et la classe du document. Les expérimentations, conduites sur le corpus DECODA, montrent que la représentation proposée permet un gain de plus de 2.5 points en termes de conversations correctement classifiées.