A Modern Greek readability tool: Development of evaluation methods

George Mikros, Rania Voskaki

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

2 Citations (Scopus)

Abstract

The aim of this paper is to develop an automatic readability analysis tool that focusses on Modern Greek as a foreign language. Based on previous work done in the Centre for the Greek Language (CGL), we offer an enhanced methodology in readability prediction for Modern Greek texts matching the adequacy level (A1 to C2) according to the Common European Framework of Languages. The proposed tool is based on several stylometric indices inspired by work done in the field of quantitative linguistics. The resulting feature vectors train a Random Forest, a robust and accurate machine learning algorithm that predicts readability in our testing dataset with 0.943 accuracy, surpassing all previous readability tools for Modern Greek. Further, analysis of the results with advanced visualization methods reveals the complex and fluid dynamics of the features used and their readability predictions.

Original languageEnglish
Title of host publicationLanguage and Text. Data, models, information and applications
EditorsAdam Pawlowski, Jan Ma?cutek, Sheila Embleton, George Mikros
PublisherJohn Benjamins Publishing Company
Pages163-175
Number of pages13
ISBN (Electronic)9789027258380
DOIs
Publication statusPublished - 2021

Publication series

NameCurrent Issues in Linguistic Theory
Volume356
ISSN (Print)0304-0763

Keywords

  • Annotation
  • Corpora
  • Evaluation methods
  • Readability tool

Fingerprint

Dive into the research topics of 'A Modern Greek readability tool: Development of evaluation methods'. Together they form a unique fingerprint.

Cite this