Shaer-AI/shaer-eval-raw-ashaar-model-300-split
收藏资源简介:
该数据集是一个用于评估AI模型在诗歌生成任务中表现的数据集,重点关注阿拉伯语诗歌(ashaar)的生成和评估。数据集包含多个特征字段,如生成ID、模型信息(模型名称、显示名称、ID、组别和角色)、文本内容(参考前缀、参考完成、生成文本等)、评估指标(如流畅度得分、连贯性得分、诗歌性得分、描述依从性得分)和韵律相关数据(如基本韵律、形式、韵律标签、韵律评估状态)。此外,数据集还包含健康状态、解码配置和时间戳等元数据。数据以测试分割形式提供,包含300个示例,用于分析和比较不同模型在诗歌生成任务中的性能,特别是韵律遵守和文本质量方面。
This dataset is designed to evaluate the performance of AI models in poetry generation tasks, with a primary focus on the generation and evaluation of Arabic poetry (ashaar). It includes multiple feature fields, such as generation ID, model information (including model name, display name, ID, group and role), text content (including reference prefix, reference completion, generated text, etc.), evaluation metrics (e.g., fluency score, coherence score, poeticity score, and description adherence score), as well as prosody-related data (such as basic prosody, form, prosody label, and prosody evaluation status). Additionally, the dataset also contains metadata including health status, decoding configuration and timestamps. The data is provided as a test split containing 300 examples, which are used to analyze and compare the performance of different models in poetry generation tasks, especially in terms of prosody compliance and text quality.




