RI: Small: Collaborative Research: Computational Methods for Argument Mining: Extraction, Aggregation, and Generation

RI：小型：协作研究：参数挖掘的计算方法：提取、聚合和生成

• 批准号: 2100885
• 负责人: Lu Wang
• 金额: $ 16.05万
• 依托单位: Regents of the University of Michigan - Ann Arbor
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-08-31 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2100885&HistoricalAwards=false
• 关键词: RI; Small; Collaborative; Research; Computational

Understanding, evaluating and generating arguments are all crucial elements in the decision-making and reasoning process. Not surprisingly then, a multitude of arguments are encountered and constructed on a daily basis as decisions are made at work and at home, in our social life and in our civic life. In spite of their ubiquity in our lives, most people are not particularly skilled in the interpretation or generation of arguments. At best, making sense of the often massive amount of argumentative online text on a topic of interest remains a daunting task. And while numerous tools exist for representing, modeling and visualizing arguments and argumentative discussions, they are limited by the substantial human effort required to input, organize and annotate arguments for use by the tools. Thus there exists a pressing need for, and this project aims to develop, automated techniques from the field of Natural Language Processing to support all facets of argumentation. This project will have a wide array of broader impacts, including providing otherresearchers with annotated datasets and tools for the analysis and generation of arguments, enhancing education through graduate and undergraduate mentoring, and promoting STEM education diversity through programs for middle and high school girls. This project aims to break new ground in the burgeoning area of argument mining. It develops a collection of computational models that comprise the basis of an argumentation toolkit---methods that can be combined and reused to support a range of argumentation applications. The project focuses on inter-related threads of research covering three critical areas of exploration for computational argumentation: (1) argument extraction---making sense of argumentative text. Drawing upon recent developments in structured learning, techniques are developed to identify the components and the structure of an argument within a single document or single turn in an online dialog. (2) Argument aggregation---clustering the components of argumentative text (e.g. sentences, turns) drawn from multiple documents according to the facets of the topic under discussion that they address. Representation learning methods are proposed to better capture topical content and argumentative styles. (3) Argument generation---constructing coherent arguments via rewriting. A neural argument generation framework with key phrase extraction as an intermediate representation is created to improve interpretation of sentences from different sources. A discourse-aware neural generation model is also investigated as an extension to improve the coherence of the generated text.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

理解、评估和产生论点都是决策和推理过程中的关键要素。因此，在工作和家庭、社交生活和公民生活中做出决定的过程中，每天都会遇到和构建大量的争论，这就不足为奇了。尽管它们在我们的生活中无处不在，但大多数人并不是特别擅长解释或生成论点。在最好的情况下，理解关于感兴趣的话题的大量在线辩论文本仍然是一项艰巨的任务。虽然有许多工具用于表示、建模和可视化论点和议论讨论，但它们受到输入、组织和注释论点以供工具使用所需的大量人力的限制。因此，迫切需要自然语言处理领域的自动化技术来支持论证的方方面面，该项目的目标是开发该技术。该项目将产生一系列更广泛的影响，包括为其他研究人员提供带注释的数据集和工具，用于分析和生成论点，通过研究生和本科生指导来加强教育，并通过针对初中和高中女孩的计划促进STEM教育的多样性。这个项目的目的是在新兴的论点挖掘领域开辟新的天地。它开发了一组计算模型，这些模型构成了论证工具包的基础-可以组合和重复使用的方法，以支持一系列论证应用程序。该项目集中于相互关联的研究线索，涵盖了计算机辩论的三个关键探索领域：(1)论点提取-使辩论文本有意义。借鉴结构化学习的最新发展，开发了在在线对话中识别单个文档或单个话轮中论点的组成部分和结构的技术。(2)论据聚合-根据所讨论的主题的方面，对来自多个文档的论辩文本的组成部分(例如句子、话轮)进行聚类。为了更好地捕捉话题内容和辩论风格，提出了表征学习方法。(3)论元生成-通过重写构建连贯的论据。建立了一个以关键短语提取为中间表示的神经论元生成框架，以提高对不同来源句子的解释能力。语篇感知的神经生成模型也被研究为改进生成文本的连贯性的扩展。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Sentence-Level Content Planning and Style Specification for Neural Text Generation

• DOI: 10.18653/v1/d19-1055
• 发表时间: 2019-09
• 期刊: ArXiv
• 作者: Xinyu Hua;Lu Wang

Argument Mining for Understanding Peer Reviews

• DOI: 10.18653/v1/n19-1219
• 发表时间: 2019-03
• 期刊: arXiv: Learning
• 作者: Xinyu Hua;M. Nikolov;Nikhil Badugu;Lu Wang

PAIR: Planning and Iterative Refinement in Pre-trained Transformers for Long Text Generation

• DOI: 10.18653/v1/2020.emnlp-main.57
• 发表时间: 2020-10
• 期刊: ArXiv
• 作者: Xinyu Hua;Lu Wang

DYPLOC: Dynamic Planning of Content Using Mixed Language Models for Text Generation

• DOI: 10.18653/v1/2021.acl-long.501
• 发表时间: 2021-06
• 期刊: ArXiv
• 作者: Xinyu Hua;Ashwin Sreevatsa;Lu Wang

Neural Argument Generation Augmented with Externally Retrieved Evidence

• DOI: 10.18653/v1/p18-1021
• 发表时间: 2018-05
• 作者: Xinyu Hua;Lu Wang-