Reading LEMI: A Corpus-Based Tool to Support Literacy in an Under-Resourced Language

Abstract

LEMI (Lectură pentru mine, ’Reading for Me’) is a corpus-based literacy support platform for Romanian, an under-resourced language in which nearly 42% of students aged 6–15 are classified as functionally illiterate despite a general literacy rate of 99%. The platform combines a curated repository of 250 children’s literary texts with an automatic readability analysis tool based on four interpretable linguistic parameters: Average Sentence Length (ASL), Percentage of Complex Words (PCW), Unique Content Word Density (UCWD), and Word Diversity (WD), combined through a weighted scoring formula designed to avoid the syllable-based biases of English readability metrics and better suited to the morphological complexity of Romanian. This paper presents LEMI’s design and documents its application across five empirical contexts: classroombased validation with primary school students, linguistic complexity assessment of school textbooks, contrastive analysis of readability in translated versus original children’s literature, lexical modernization of canonical literary texts, and in-service teacher training for corpus-informed text adaptation. Across these contexts, LEMI has functioned as a diagnostic, benchmarking, and decision-support tool, producing interpretable outputs that complement rather than replace professional judgment. The platform adheres to FAIR data principles and implements a metadata schema supporting federated content search and reuse across European research infrastructures. By making both tools and data openly available, LEMI contributes a transparent, reusable resource for readability research and literacy support in an under-resourced language.