QCRI/Fanar-Sadiq-Bench
收藏资源简介:
Fanar-Sadiq分类器数据集是一个双语阿拉伯语/英语的查询分类数据集,随论文Fanar-Sadiq: A Multi-Agent Architecture for Grounded Islamic QA发布。该数据集支持Fanar-Sadiq多代理伊斯兰问答系统的路由组件,该系统将用户查询路由到专门模块,如基于检索的伊斯兰问答、古兰经检索、祈祷词查找、祈祷时间计算、伊斯兰日历工具、天课计算器和继承计算器。数据集包含两个分类器子集:提示分类器(用于决定用户查询是否与伊斯兰相关)和意图分类器(用于将伊斯兰查询路由到适当的专门组件)。数据包括阿拉伯语和英语查询,涵盖自然表达、代码切换、简短、冗长、事实性、法学性、对话式和工具导向的用户请求。
The Fanar-Sadiq Classifier Dataset is a bilingual Arabic/English query classification dataset accompanying the paper titled Fanar-Sadiq: A Multi-Agent Architecture for Grounded Islamic QA. This dataset supports the routing component of the Fanar-Sadiq multi-agent Islamic question answering (QA) system, which directs user queries to specialized modules including retrieval-based Islamic QA, Quran retrieval, supplication lookup, prayer time calculation, Islamic calendar tool, Zakat calculator, and inheritance calculator. The dataset comprises two classifier subsets: the prompt classifier (for determining whether a user query is Islamic-related) and the intent classifier (for routing Islamic queries to appropriate specialized components). The dataset includes Arabic and English queries covering natural expressions, code-switching, short, lengthy, factual, jurisprudential, conversational, and tool-oriented user requests.




