An empirical evaluation of Lexiscore®'s predictive validity in assessing Greek high school writing

This study evaluates the predictive capacity of Lexiscore®, a rule-based electronic assessment tool designed for the automated evaluation of student writing in Greek. Specifically, we investigated how well selected linguistic features extracted by the system predict human-assigned scores across three key parameters of written discourse: content, organization, and language use. Using multiple regression analyses, we tested the relationship between these features and teacher evaluations of high school essays. The results demonstrate that Lexiscore® can explain a substantial proportion of variance in human scoring—25.05% for content, 20.31% for organization, and 37.14% for language—highlighting its potential as a reliable tool for supporting classroom-based writing assessment. Limitations related to language resources and NLP tools for Greek are discussed, along with implications for future development and pedagogical integration.