NYT Book Opinions
收藏资源简介:
NYT Book Opinions数据集由斯坦福大学创建,旨在研究语言模型对非政治主观意见的响应。该数据集包含235本过去二十年中由《纽约时报》评选的热门书籍的评价,346名标注者提供了4点Likert评分。数据集的创建过程涉及收集书籍信息、标注者评分及统计分析。该数据集主要用于评估语言模型在模拟人类阅读兴趣分布方面的能力,旨在解决语言模型在非政治领域的主观意见表达问题。
The NYT Book Opinions Dataset was developed by Stanford University for investigating language models' responses to non-political subjective opinions. This dataset includes 4-point Likert scale ratings from 346 annotators, covering reviews of 235 popular books selected by The New York Times over the past two decades. The dataset construction process involves collecting book-related information, annotator-provided ratings, and conducting statistical analyses. It is primarily used to evaluate the ability of language models to simulate the distribution of human reading interests, aiming to address the issue of subjective opinion expression by language models in non-political domains.




