{"id":1262,"date":"2024-07-21T17:09:06","date_gmt":"2024-07-21T19:09:06","guid":{"rendered":"https:\/\/sites.usp.br\/keml\/?page_id=1262"},"modified":"2024-08-09T10:40:26","modified_gmt":"2024-08-09T12:40:26","slug":"argument-mining","status":"publish","type":"page","link":"https:\/\/sites.usp.br\/keml\/en\/argument-mining\/","title":{"rendered":"Argument Mining"},"content":{"rendered":"
<\/h6>\n
\"\"<\/a>

AI-generated image (Copilot)<\/p><\/div>\n

One of the traits of rational behavior is the ability to provide justifications for actions and beliefs. Such justifications are organized in the form of arguments: discursive structures that support or attack a conclusion that is not immediately obvious. Arguments are the main resource for explaining our decisions, convincing others of something we believe, and producing consensual agreements.<\/p>\n

The study of argumentation constitutes a traditional field of artificial intelligence (AI), dating back at least to the 1980s. Most of the work developed involves the construction of formal systems for building and evaluating arguments. There are at least two major tasks: argument mining and argument reasoning\/evaluation. More recently, with the increasing advancement of large language models, such as GPT, there has been an interest in reconciling the linguistic abilities and world knowledge of these language models with the ability to argue rigorously.<\/p>\n

KEML has a team dedicated to the study of argumentation models in AI. The idea is to produce models capable of identifying and evaluating arguments written in natural language. The main branches:<\/p>\n

– Training encoder models for argument mining tasks.
\n– Studying argumentation semantics.
\n– Argumentation based on large language models.<\/p>\n

Learn a bit more about \u2026.<\/p>\n