The automatic assessment of text readability and the classification of texts by levels is essential for language education and language-related industries that rely on effective communication. In European Portuguese, most of the studies on this subject focus on identifying the level of texts used for proficiency evaluation purposes according to the Common European Framework of Reference for Languages (CEFR). However, the ordinal nature of the levels is not considered by the classification models used in those studies. In this paper, we address the problem as a regression task in an attempt to leverage that information. Our experiments using fine-tuned versions of a state-of-the-art foundation model for Portuguese show that addressing the problem as a regression task leads to improved performance in terms of adjacent accuracy and improved generalization ability to different kinds of textual data.

--

Esta comunicação está associada ao registo seguinte:

• Text Readability Assessment in European Portuguese: A Comparison of Classification and Regression Approaches