ArBanking77

ArBanking77是一个大型阿拉伯语数据集，专注于银行领域的意图检测。该数据集由比尔宰特大学的研究团队从原始的英文Banking77数据集本地化和阿拉伯化而来，包含31,404条查询，涵盖现代标准阿拉伯语（MSA）和巴勒斯坦方言，每条查询被分类为77个类别之一。数据集的创建过程包括半自动翻译和手动验证修订，旨在通过模拟低资源设置和噪声查询来增强模型的鲁棒性。该数据集主要用于训练基于AraBERT的神经模型，以解决银行领域的意图检测问题。ArBanking77 is a large-scale Arabic dataset dedicated to intent detection in the banking domain. Developed by a research team from Birzeit University via localization and Arabization of the original English Banking77 dataset, it contains 31,404 queries covering both Modern Standard Arabic (MSA) and Palestinian dialects, with each query assigned to one of 77 predefined categories. The dataset construction process adopts semi-automatic translation combined with manual verification and revision, aiming to enhance model robustness by simulating low-resource scenarios and noisy queries. This dataset is primarily utilized for training AraBERT-based neural models to solve intent detection tasks in the banking sector.