freelawproject/opinions-synthetic-query-8192

这个数据集用于对语义搜索模型进行微调，它包含chunk_id、anchor、positive和negative四个字段，其中chunk的最大长度为7800个token，使用bert-base-cased分词器，并且有2个sentence的overlap。数据集分为训练集和开发集，分别有351和95个样本。该数据集适用于句子相似度任务。This dataset is used for fine-tuning semantic search models, it includes four fields: chunk_id, anchor, positive, and negative, where the maximum length of a chunk is 7800 tokens, tokenized using bert-base-cased tokenizer with 2 sentence overlap. The dataset is split into a training set and a development set, containing 351 and 95 samples respectively. It is suitable for sentence similarity tasks.