Exploring Artificial Intelligence in Language Teaching: Automatic Creation and Classification of Narrative Corpora Based on Proverbs for Teaching PFL with Large Language Models

Abstract

Artificial Intelligence (AI), particularly Large Language Models (LLMs), has been transforming language education. This study explores the application of LLMs in the teaching of Portuguese as a Foreign Language (PFL), focusing on the automatic creation, classification, and validation of a corpus of short narratives based on Portuguese proverbs. The objectives are: (i) to automatically generate short narratives based on proverbs, suitable for different proficiency levels as defined by the CEFR; (ii) to automatically classify the linguistic proficiency level of these narratives using traditional machine learning techniques and foundational models (LLMs), followed by validation by human evaluators; (iii) to assess the adequacy of the narratives in relation to the original proverbs, justifying their potential didactic use in the context of PFL teaching. The methodology involves generating narratives with LLMs, which are then validated using automatic tools and human experts, as well as analysing the correspondence between the proverb and the generated narrative with a view to pedagogical application.