地球资源数据云——数据资源详情

人工智能与人类科学文本数据集 (AIGTxt)

发布时间:2026-03-17 14:32:26资源ID:2031260809790132226资源类型:免费

该数据集《AI vs Human Scientific Text Dataset (AIGTxt)》主要用于监督学习任务,数据形态以文本为主,应用场景偏向文本内容分析。 题目说明:Collection of human, AI, and mixed scientific texts for detection research 任务类型:文本监督学习。 建议流程:先做文本清洗与分词,再比较 TF - IDF+线性模型 与 预训练语言模型。 评估建议:使用分层切分或交叉验证,优先关注 F1、Recall、AUC 等分类指标。 可用文件:未检测到标准 CSV,可优先查看目录中的索引或说明文件。 he AI vs Human Scientific Text Dataset (AIGTxt) is created to support research on detecting AI - generated scientific writing. It contains 10,821 paragraphs across ten domains, including Computer Science, Medicine, Social Sciences, and more. Context & Inspiration: With AI models like ChatGPT producing highly human - like text, distinguishing AI - written content from human writing has become crucial. This dataset was inspired by the need to study linguistic and stylistic differences and to benchmark AI text detection methods. Sources & Composition: Human - written texts: Extracted from academic papers (Introduction, Background, or Literature Review sections). ChatGPT - generated texts: Manually rephrased versions of the human texts.

人工智能与人类科学文本数据集 (AIGTxt)

摘要概览

该数据集《AI vs Human Scientific Text Dataset (AIGTxt)》主要用于监督学习任务,数据形态以文本为主,应用场景偏向文本内容分析。 题目说明:Collection of human, AI, and mixed scientific texts for detection research

任务类型:文本监督学习。

建议流程:先做文本清洗与分词,再比较 TF - IDF+线性模型 与 预训练语言模型。

评估建议:使用分层切分或交叉验证,优先关注 F1、Recall、AUC 等分类指标。

可用文件:未检测到标准 CSV,可优先查看目录中的索引或说明文件。

he AI vs Human Scientific Text Dataset (AIGTxt) is created to support research on detecting AI - generated scientific writing. It contains 10,821 paragraphs across ten domains, including Computer Science, Medicine, Social Sciences, and more.

Context & Inspiration: With AI models like ChatGPT producing highly human - like text, distinguishing AI - written content from human writing has become crucial. This dataset was inspired by the need to study linguistic and stylistic differences and to benchmark AI text detection methods.

Sources & Composition:

Human - written texts: Extracted from academic papers (Introduction, Background, or Literature Review sections).

ChatGPT - generated texts: Manually rephrased versions of the human texts.