HANNA
收藏资源简介:
HANNA数据集是由巴黎综合理工学院电信学院创建的,包含1056个由10种不同的自动故事生成系统生成的故事。每个故事都与一个人类故事相关联,并由3个不同的评价者根据6个人类评价标准进行标注。该数据集旨在量化评估自动评价指标与人类评价标准之间的相关性,特别关注故事生成的质量和创造性。HANNA数据集的应用领域包括游戏、通信和教育,旨在通过标准化和广泛的人类评价来加强故事生成的评估。
The HANNA dataset was developed by the School of Telecommunications of École Polytechnique (Paris). It contains 1056 stories generated by 10 distinct automatic story generation systems, with each story paired with a corresponding human-written counterpart. Each story is annotated by three independent evaluators based on six human evaluation criteria. This dataset aims to quantitatively assess the correlation between automatic evaluation metrics and human evaluation standards, with a particular focus on the quality and creativity of story generation. The HANNA dataset has applications in fields including games, communications and education, and is designed to enhance the evaluation of story generation via standardized and extensive human evaluations.

- 1Of Human Criteria and Automatic Metrics: A Benchmark of the Evaluation of Story Generation巴黎综合理工学院电信学院 · 2022年



