jhu-clsp/mFollowIR-parquet

mFollowIR数据集的parquet版本，可以直接使用`load_dataset()`加载。该数据集包含多种语言的配置，每种语言都有原始和修改后的相关性判断、文档集合、查询集以及排名靠前的文档。数据集的结构包括`qrels_og_[lang]`、`qrels_changed_[lang]`、`corpus_[lang]`、`queries_[lang]`和`top_ranked_[lang]`等配置。

The mFollowIR-parquet dataset is a parquet version of the mFollowIR dataset, which can be loaded directly using `load_dataset()`. The dataset contains configurations for three languages: Farsi (fas), Russian (rus), and Chinese (zho). Each language has five configurations: original relevance judgments (qrels_og_[lang]), modified relevance judgments (qrels_changed_[lang]), document collection (corpus_[lang]), query set with instructions (queries_[lang]), and top ranked documents (top_ranked_[lang]).