Shaer-AI/shaer-sft-test-generations-k5-meter-count
收藏资源简介:
该数据集名为Shaer SFT Test Generations K=5 Meter/Count Scores,是基于Shaer-AI/shaer-sft-test-generations-k5数据集增强而来的,专门添加了自动韵律和请求行数评估功能。数据集包含17,405行数据,使用4BiLSTM韵律分类器和行数依从性进行度量评估,核心指标列包括meter(韵律)、count_adherence(行数依从性)、parsed_num_lines(解析行数)、requested_num_lines(请求行数)、meter_eval_status(韵律评估状态)和count_eval_status(行数评估状态)。整体韵律平均值为0.6520384872952881,行数依从性平均值为0.9783539094873928。自动韵律评分主要用于聚合评估和错误分析,不能替代专家对选定诗歌的手动审查。数据集涉及阿拉伯语诗歌(Arabic poetry)和韵律评估(meter evaluation),任务类别为文本生成(text-generation)。
This dataset, titled Shaer SFT Test Generations K=5 Meter/Count Scores, enriches the Shaer-AI/shaer-sft-test-generations-k5 dataset by adding automatic meter and requested-line-count evaluation. It contains 17,405 rows and uses a 4BiLSTM meter classifier plus line-count adherence for metric evaluation. Core metric columns include meter, count_adherence, parsed_num_lines, requested_num_lines, meter_eval_status, and count_eval_status. The overall meter mean is 0.6520384872952881, and the overall count-adherence mean is 0.9783539094873928. The automatic meter score is intended for aggregate evaluation and error analysis and is not a substitute for expert manual review of selected poems. The dataset is related to Arabic poetry and meter evaluation, with task categories in text-generation.




