MarcGrumpyOlejak/gooaq_mt_german_5_hard_negatives
收藏Hugging Face2025-07-30 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/MarcGrumpyOlejak/gooaq_mt_german_5_hard_negatives
下载链接
链接失效反馈官方服务：
资源简介：
GooAQ德语数据集是一个包含约200万个问题-答案-负样本三元组和问题-答案-负样本_1...-负样本_5元组的数据集，适用于训练嵌入模型。负样本是通过静态双语sts-mrl-en-de-base-v1模型自动挖掘得到的，每个问题-答案对挖掘5个负样本。该数据集可以直接与Sentence Transformers一起使用。
The GooAQ German dataset consists of approximately 2 million question-answer-negative triplet and question-answer-negative_1...-negative_5 tuples suitable for training embedding models. The negative samples are automatically mined using the static bilingual sts-mrl-en-de-base-v1 model, with 5 negatives mined per question-answer pair. This dataset can be used directly with Sentence Transformers.
提供机构：
MarcGrumpyOlejak
