agentlans/sentence-paraphrases

句子重写数据集是一个经过精心策划的句子长度重写句子的集合，主要来源于两个数据集：humarin的ChatGPT-paraphrases和xwjzds的paraphrase_collections。该数据集的结构旨在提供原始文本及其重写句子的成对句子。每个条目包括一个文本字段，其中包含最不直观的重写句子，以及一个重写文本字段，其中包含最直观的重写句子。直观性是通过agentlans/deberta-v3-xsmall-zyda-2-readability模型评估的。在处理过程中，移除了重复的行和不正确的输出，并根据直观性分数选择了重写句子。

This dataset is a curated collection of sentence-length paraphrases derived from two primary sources: humarins ChatGPT-paraphrases and xwjzdss paraphrase_collections. The dataset is structured to provide pairs of sentences from an original text and its paraphrase(s). Each entry includes a text field containing the least readable paraphrase and a paraphrase field containing the most readable paraphrase, assessed using the agentlans/deberta-v3-xsmall-zyda-2-readability model. During processing, duplicate rows and incorrect outputs were removed, and paraphrases were selected based on readability scores.