ReaderBench: The Learning Companion

Mihai Dascalu (1) , Larise L Stavarache (2) , Philippe Dessus (3) , Stefan Trausan-Matu (1) , Danielle A Mcnamara (4) , Maryse Bianco (3)

1 Université Politehnica
2 Computer Science Department
3 LSE - Laboratoire des Sciences de l'Éducation (Grenoble)
4 ASU - Arizona State University [Tempe]

Continuous progress tracking in terms of automated essay scoring, assessment of reading strategies, and evaluation of learners' involvement in collaboration groups represents a key component in technology-scaffolded learning. Our educational software, ReaderBench [1, 2], is based on current research in the automated essay scoring field (E-rater, iSTART, Coh-Metrix), but provides an integrated approach centered on cohesion. ReaderBench supports both tutors and students, affording automated evaluations of reading strategies, course materials selection, and CSCL collaboration. ReaderBench has been designed to flexibly allow multiple configurations for various educational scenarios and languages (English, French, and Italian). 1 ReaderBench's Purpose ReaderBench targets both tutors and students by providing a fully functional learning model approach including invidual and collaborative learning methods, cohesion-based discourse analysis [2], dialogical discourse model [3], textual complexity evaluation [1], reading strategies identification [4], and participation and collaboration assessment [5]. By using natural language processing techniques, the main purpose of this framework is to bind traditional learning methods with new trends and technologies to support computer supported collaborative learning (CSCL). ReaderBench, by design, is not meant to replace the tutor, but to scaffold both tutors and learners by enabling continuous assessment, self-assessment, collaborative evaluation of individuals' contributions, as well as the analysis of reading materials to match readers to their appropriate class level text. Overall, ReaderBench is a fully functional automated software framework, designed to be an educational helper for students and tutors. The system makes uses of text-mining techniques based on advanced natural language processing and machine learning algorithms to design and deliver summative and formative assessments using multiple data sets (e.g., textual materials, behavior tracks, self-explanations).

Mots clés: Natural Language Processing

Domaines: Education; Environnements Informatiques pour l'Apprentissage Humain

Mihai Dascalu, Larise L Stavarache, Philippe Dessus, Stefan Trausan-Matu, Danielle A Mcnamara, et al.. ReaderBench: The Learning Companion. 17th Int. Conf. on Artificial Intelligence in Education (AIED 2015), 2015, Madrid, Spain. ⟨hal-01672087⟩