Argumentation is a fundamental aspect of our language and communication which involves the use of reasoning, to justify our opinions and persuade others about them. A good argumentation is fundamental in drawing widely accepted conclusions and not only for decision making and learning. It is employed in various contexts including debates, discussions, essays, legal proceedings, and everyday conversations. Understanding the argumentative structure of a discourse makes it possible to determine not only which are the positions adopted, but also giving reasons for them. In our research we consider a widely used argumentative context which is the one of argumentative essays. These are a type of academic writing that presents a stance on a particular issue and supports it with evidence and reasoning. The purpose of an argumentative essay is to persuade the reader to accept or agree with the writer’s point of view on a controversial or debatable topic. The essay typically follows a specific structure and involves the systematic development of an argument. Using a famous and widely studied corpus of university student essays —the Argument Annotated Essays Corpus (AAEC)—, we performed the tasks of argument identification and classification using a mix of text data augmentation techniques, also taking advantage of the new potentials made available by large language models. After an extensive introduction and survey to artificial intelligence, machine learning, argumentation theories and computational argumentation, which aim at providing a solid background for the rest of the work, we present our own contribution to the field of argument mining, experimentally showing that our techniques are able to measurably go beyond the state of the art for these tasks in AAEC.
ARGUMENTATION MINING OF ARGUMENTATIVE ESSAYS(2025 Jan 15).
ARGUMENTATION MINING OF ARGUMENTATIVE ESSAYS
DEMARIA, ROBERTO
2025-01-15
Abstract
Argumentation is a fundamental aspect of our language and communication which involves the use of reasoning, to justify our opinions and persuade others about them. A good argumentation is fundamental in drawing widely accepted conclusions and not only for decision making and learning. It is employed in various contexts including debates, discussions, essays, legal proceedings, and everyday conversations. Understanding the argumentative structure of a discourse makes it possible to determine not only which are the positions adopted, but also giving reasons for them. In our research we consider a widely used argumentative context which is the one of argumentative essays. These are a type of academic writing that presents a stance on a particular issue and supports it with evidence and reasoning. The purpose of an argumentative essay is to persuade the reader to accept or agree with the writer’s point of view on a controversial or debatable topic. The essay typically follows a specific structure and involves the systematic development of an argument. Using a famous and widely studied corpus of university student essays —the Argument Annotated Essays Corpus (AAEC)—, we performed the tasks of argument identification and classification using a mix of text data augmentation techniques, also taking advantage of the new potentials made available by large language models. After an extensive introduction and survey to artificial intelligence, machine learning, argumentation theories and computational argumentation, which aim at providing a solid background for the rest of the work, we present our own contribution to the field of argument mining, experimentally showing that our techniques are able to measurably go beyond the state of the art for these tasks in AAEC.File | Dimensione | Formato | |
---|---|---|---|
PhD_Thesis_ROBERTO_DEMARIA.pdf
Accesso aperto
Descrizione: Tesi
Dimensione
6.92 MB
Formato
Adobe PDF
|
6.92 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.