Scholarlink

<p>The use of quantum computing to analyze text is still at a nascent stage, especially when dealing with low-resource languages like Arabic, which has a notably intricate structure. The current state of quantum technology still limits the development of powerful models for processing Arabic text. This paper introduces AraBERT-QC, a novel hybrid model that integrates the Arabic language model AraBERT with a quantum classifier. We leverage AraBERT to generate contextualized sentence embeddings that capture the semantic structure of Arabic text, which are subsequently processed by the quantum classifier. Our experimental study is structured in two parts. The first part investigates the impact of various quantum data embedding techniques on the accuracy and efficiency of the quantum support vector machine (QSVM) algorithm. The second part explores the performance and efficiency trade-offs between two hybrid quantum-classical approaches: Hybrid quantum-classical neural networks (H-QNN) and the variational quantum classifier (VQC). We evaluate our architecture on two distinct datasets, sarcasm detection and sentiment analysis, using metrics such as accuracy, macro <i>F</i>1-score, precision, recall, and computation time. Results show that complex quantum embeddings achieve better performance in some tasks, while the VQC trains faster than the H-QNN. Notably, AraBERT-QC achieves 86.25% accuracy in sarcasm detection and 84.33% in sentiment analysis, outperforming the classical SVM and neural network baselines by up to 2% in accuracy and <i>F</i>1-score. This study highlights the potential of integrating quantum techniques with classical models to tackle challenges in computational linguistics, particularly for morphologically rich languages.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AraBERT-QC: a novel quantum-based classification architecture to classify short Arabic sentences

  • 期刊论文

  • 发表日期：

    2025年10月15日

查看全文
  • Islam Djemmal,
  • Hacene Belhadef

摘要

The use of quantum computing to analyze text is still at a nascent stage, especially when dealing with low-resource languages like Arabic, which has a notably intricate structure. The current state of quantum technology still limits the development of powerful models for processing Arabic text. This paper introduces AraBERT-QC, a novel hybrid model that integrates the Arabic language model AraBERT with a quantum classifier. We leverage AraBERT to generate contextualized sentence embeddings that capture the semantic structure of Arabic text, which are subsequently processed by the quantum classifier. Our experimental study is structured in two parts. The first part investigates the impact of various quantum data embedding techniques on the accuracy and efficiency of the quantum support vector machine (QSVM) algorithm. The second part explores the performance and efficiency trade-offs between two hybrid quantum-classical approaches: Hybrid quantum-classical neural networks (H-QNN) and the variational quantum classifier (VQC). We evaluate our architecture on two distinct datasets, sarcasm detection and sentiment analysis, using metrics such as accuracy, macro F1-score, precision, recall, and computation time. Results show that complex quantum embeddings achieve better performance in some tasks, while the VQC trains faster than the H-QNN. Notably, AraBERT-QC achieves 86.25% accuracy in sarcasm detection and 84.33% in sentiment analysis, outperforming the classical SVM and neural network baselines by up to 2% in accuracy and F1-score. This study highlights the potential of integrating quantum techniques with classical models to tackle challenges in computational linguistics, particularly for morphologically rich languages.