ArgKP-2021

该数据集基于 ArgKP 数据集，其中包含人群就 28 个有争议的主题贡献的论点，按他们对该主题的立场划分，以及专家为这些主题编写的 KP。数据集涵盖了一组有争议的主题，其中每个主题和立场，一组三元组，形式为<argument、KP、label>提供。 . 收集人群注释以确定 KP 是否代表一个论点，即是否与论点匹配。 ArgKP 中的参数是 IBM-ArgQ-Rank-30kArgs 数据集的子集。对于测试集，我们扩展了 ArgKP，添加了三个新的有争议的主题，它们也不属于 IBM-ArgQ-Rank-30kArgs。该测试集是专门为 KPA-2021 收集的，并且经过精心设计，在各个方面都与训练数据 2 相似。对于每个主题，收集众包参数，生成专家 KP，并获得参数/KP 对的匹配/不匹配注释，从而生成与 ArgKP 格式兼容的数据集。参数收集严格遵守 IBM-ArgQ-Rank-30kArgs 中用于收集参数的准则、质量度量和后处理，而专家 KP 的生成、匹配注释的收集和最终数据集的创建严格遵守创建 ArgKP 的方式。

This dataset is based on the ArgKP dataset, which contains arguments contributed by crowdworkers on 28 controversial topics, grouped by their stances toward the topics, along with KPs authored by experts for these topics. The dataset covers a set of controversial topics, where for each topic and stance, a set of triplets in the format of <argument, KP, label> is provided. Crowdsourcing annotations were collected to determine whether a KP represents an argument, i.e., whether the KP matches the given argument. The arguments included in ArgKP constitute a subset of the IBM-ArgQ-Rank-30kArgs dataset. For the test set, we extended ArgKP by adding three new controversial topics that also do not belong to the IBM-ArgQ-Rank-30kArgs dataset. This test set was specifically collected for KPA-2021, and was carefully designed to be highly similar to training data 2 across all aspects. For each topic, crowd-sourced arguments were collected, expert KPs were generated, and match/mismatch annotations were obtained for argument-KP pairs, resulting in a dataset compatible with the ArgKP format. The collection of arguments strictly follows the guidelines, quality metrics, and post-processing procedures used for argument collection in the IBM-ArgQ-Rank-30kArgs dataset, while the generation of expert KPs, collection of match annotations, and creation of the final dataset strictly adhere to the methodology employed to develop the original ArgKP dataset.

搜集汇总

背景与挑战

背景概述

ArgKP-2021是一个包含争议主题论点与关键点匹配标注的数据集，扩展了测试集并严格遵循了数据收集和处理标准。