WikiBioCTE

WikiBioCTE 是基于现有数据集 WikiBio（最初为表格到文本生成而创建）的可控文本编辑数据集。在可控文本编辑任务中，输入是长文本、问题和目标答案，输出是经过最小修改的文本，使其适合目标答案。这项任务在许多情况下都非常重要，例如更改法律文件中的某些条件、后果或属性，或者更改新闻文本中事件的某些关键信息。WikiBioCTE is a controllable text editing dataset derived from the existing WikiBio corpus, which was initially developed for table-to-text generation tasks. For a controllable text editing task, the inputs include long-form text, a query and a target answer, while the output is a minimally modified text that aligns with the target answer. This task is critically important in numerous scenarios, such as revising certain conditions, consequences or attributes in legal documents, or altering key information about events in news articles.

搜集汇总

背景与挑战

背景概述

WikiBioCTE是一个基于WikiBio构建的可控文本编辑数据集，用于通过最小修改长文本来适配目标答案，适用于法律文件或新闻文本编辑等场景。该数据集由牛津大学于2021年发布。