mysamai/ashaar-tashkeel

Ashaar — Model-Inferred Tashkeel数据集是一个包含6,934,210个阿拉伯诗歌诗句的模型推断音标化（tashkeel）数据集。数据来源于arbml/ashaar数据集，并通过basharalrfooh/Fine-Tashkeel模型进行音标化处理。该数据集的主要用途是作为الشاعر (al-Shaaer)阿拉伯诗歌AI系统的一部分，用于引导qafiya（押韵）规则挖掘。数据集的局限性包括音标化质量是模型推断的，未经人工验证，且在处理罕见诗歌形式、方言输入或模糊的词尾元音时，错误分布未进行详细描述。

The Ashaar — Model-Inferred Tashkeel dataset consists of model-inferred diacritization (tashkeel) for 6,934,210 Arabic poetry verses. The data is sourced from the arbml/ashaar dataset and processed using the basharalrfooh/Fine-Tashkeel model. The primary use of this dataset is as part of the الشاعر (al-Shaaer) Arabic poetry AI system to bootstrap qafiya (rhyme) rule mining. Limitations of the dataset include the diacritization quality being model-inferred and not hand-verified, with no detailed characterization of error distribution on rare poetic forms, dialectal input, or ambiguous word-final vowels.