Year: 2022
Type: dissertation
Abstract: Capturing Discourse Structure and Reasoning in A瑶umentation (論 述における談話構造および論理構造の解析) 論文審査委員 (主査)東北大学教授乾健 太郎 東北大学教 授伊 藤彰則 東北大学教 授北 村喜文 東北大学教 授鈴 木潤 論文内容の要旨 Chapter 1: Introduction A 叩 mentation can help students improve their critical thinking skills, decision making skills or writing skills.However, students often struggle to construct well-defined arguments and it is necessa巧to guide them by providing feedback so that they can improve their argumentation.The difficulty here is that providing feedback manually is an extremely time consuming task and requires lots of human efforts.Therefore, the importance of building an automated feedback system is enormous and assessing the quality of argumentation and capturing its underlying reasoning patterns are two of the crucial tasks to reach this ultimate goal.Assessing the quality of argumentation can help students learn how good or bad their arguments are and the capturing the reasoning patterns can help inspect the issues that makes their arguments poor.This dissertation focuses on these two important tasks.Existing approaches for automated evaluation of argumentative texts typically rely on parsers to capture argumentative discourse.However, the performance of parsers is not always adequate, especially when they are used on noisy texts, such as student essays.To overcome this problem, in this thesis, we establish an unsupervised pre-training approach to capture argumentative discourse that does not require any parser or annotation.Our proposed unsupervised approach achieves state-of-the-art result on a task of scoring student essays.Essay scoring systems provide feedback about the quality of argumentative essays but do not indicate the issues why the quality of an essay is good or bad.In order to improve the quality of students'argumentative writing, we need feedback systems that do not only provide a score for an argumentative text, but at the same time allow students to inspect the issues、in their text.To ー
Author: Mim, Farjana Sultana
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7145748311