IAM

IAM数据集是由阿里巴巴达摩院与新加坡科技设计大学合作创建的，旨在支持一系列辩论相关的自然语言处理任务，如主张提取、立场分类和证据提取等。该数据集包含超过1000篇文章，覆盖123个话题，总计近7万条句子，每条句子都根据其辩论属性进行了详尽标注。IAM数据集不仅支持单一任务的研究，还提出了两个新的综合任务，即主张提取与立场分类（CESC）和主张-证据对提取（CEPE），以更有效地自动化辩论准备过程。这些任务的提出和数据集的创建，为辩论系统的自动化研究提供了新的视角和挑战，同时也展示了在多个实际应用中的潜力。

The IAM Dataset was co-created by Alibaba DAMO Academy and the Singapore University of Technology and Design, aiming to support a series of debate-related natural language processing tasks such as claim extraction, stance classification and evidence extraction. This dataset contains over 1,000 articles covering 123 topics, with a total of nearly 70,000 sentences, each of which has been thoroughly annotated according to its debate-related attributes. The IAM Dataset not only supports single-task research, but also proposes two novel comprehensive tasks: Claim Extraction and Stance Classification (CESC) and Claim-Evidence Pair Extraction (CEPE), to more effectively automate the debate preparation process. The proposal of these tasks and the construction of the dataset provide new perspectives and challenges for automated research on debate systems, while also demonstrating their potential in multiple practical application scenarios.

搜集汇总

背景与挑战

背景概述

IAM数据集是一个由阿里巴巴达摩院与新加坡科技设计大学合作创建的辩论相关自然语言处理数据集，包含超过1000篇文章、123个话题和近7万条句子，每条句子都根据辩论属性进行了详尽标注。它支持主张提取、立场分类和证据提取等任务，并引入了主张提取与立场分类（CESC）和主张-证据对提取（CEPE）两个综合任务，以推动辩论准备过程的自动化研究，展示出在多个实际应用中的潜力。