Un corpus di testi trilingue per promuovere la riflessione sulla pratica didattica

The purpose of this contribution is to present the Scolinter corpus, consisting of 1305 children’s texts at the end of the first primary class in three languages: French, Italian and Spanish. The research has three main objectives: firstly, it aims to show how it was possible to create a corpus using Automatic Language Processing despite the fact that the texts collected are far from the orthographic convention; secondly, it aims to describe the potential of the corpus to analyze and evaluate the spelling and the narrative competences of first grade pupils, also in a comparative way; finally, it aims to highlight how this corpus can be a starting point to observe, analyze and understand the usual practices of teaching writing, thus fostering teacher training.