2026· International Conference on Data Technologies and Applications· pp. 110-120· 0 citations· 30 references
Computer Science
TL;DR
A controlled comparison between architecture based optimization and prompting based inference for Arabic-English code-switched topic classification is provided, and the results indicate that inference time conditioning can be more effective than increasing architectural complexity for this task.
Abstract
: Arabic-English code-switching is common in online communication, where users mix both languages within the same sentence. Mixed scripts, dialectal Arabic, Arabizi, and inconsistent spelling create instability for NLP models and make tasks such as topic classification difficult. Previous work has typically focused either on improving transformer architectures through ensembles and multi view modeling, or on enhancing inference through prompt design in large language models. However, these directions are rarely evaluated under the same experimental setup. In this paper, we provide a controlled comparison between architecture based optimization and prompting based inference for Arabic-English code-switched topic classification, and we release ArEnTC , a 105k-sentence Arabic-English code-switched dataset annotated for nine topics. Architectural experiments include single model baselines, voting ensembles, stacked meta learning, and translation based multi view inference. Prompting experiments evaluate zero shot, few shot, retrieval augmented generation (RAG), and reasoning based strategies using LLaMA 3.3 70B without fine tuning. While architectural extensions progressively improve performance, reaching 0.92 Macro F1 with translation based multi view and Random Forest, retrieval augmented few shot prompting achieves 0.98 Macro F1, surpassing all ensemble configurations. The results indicate that inference time conditioning can be more effective than increasing architectural complexity for this task.
: Multilingual speakers often alternate between languages within a conversation, a phenomenon known as code-switching. This is common in Arabic-speaking communities, where Arabic and English are frequently mixed in everyday communication. Although recent advances in natural language processing have been driven by pretrained multilingual language models, these models are largely trained on monolingual data and often struggle to capture the abrupt language transitions and cross-lingual semantic interactions that characterize code-switched text. This work investigates representation learning for Arabic-English code-switched text at multiple levels. At the word level, we employ a code-switch-aware masked language modeling objective that captures token-level language variation and switch points. At the sentence level, we adopt a contrastive learning framework with natural language inference supervision to encourage semantically consistent sentence embeddings across monolingual and code-switched variants. To support this objective, we introduce CS-SNLI, a large-scale Arabic-English code-switched natural language inference dataset. The resulting embeddings are evaluated on sentiment analysis and named entity recognition. On sentiment analysis, the word-level pipeline improves F1-score by 2.17 percentage points over mBERT and 2.27 percentage points over XLM-R, while the sentence-level pipeline improves F1-score by 1.78 and 1.28 percentage points, respectively. In contrast, named entity recognition shows only marginal, statistically insignificant gains.
Mariam Rizkallah, Amani Ghonim, A. Sherif et al.· International Conference on...· 0 citations
A bidirectional Mamba encoder pretrained via masked language modeling on a corpus combining Arabic Wikipedia and CulturaX text is introduced, trained end-to-end on four consumer-grade NVIDIA RTX 2080Ti GPUs (11GB) over approximately ten days.
Ahmed Amine Aliane, H. Aliane, N. Semmar· 0 citations
An in-depth evaluation of instruction tuning for Arabic NLP tasks using three prominent LLMs: LLaMA 3.1-8B, AceGPT-v2-8B, and Qwen3-8B shows that instruction tuning consistently improves performance across most tasks, with notable variations in effectiveness across different tasks and prompts.
Maged Saeed Al-shaibani, Zaid Alyafeai, Irfan Ahmad· Language Resources and Evalu...· 0 citations
This paper proposes a general optimization framework that combines a vocabulary pruning method with a targeted fine-tuning protocol for MNMT models, and reduces the vocabulary size from over 128,000 to approximately 10,000 tokens, enabling a 60% memory saving without any loss in performance.
Ahmed Amine Aliane, N. Semmar, H. Aliane· 0 citations
The results indicate that a moderately sized, shared self-attention architecture can deliver production-quality multilin-gual translation within the resource constraints of an academic de-ployment, while surfacing clear directions – low-resource language coverage, domain adaptation, and speech-based extension – for con-tinued development.
Darshan Gowda D H and Dr. Kruti R· International Journal of Adv...· 0 citations
