This paper is from Session 5B: Methods for large scale assessment of meaningful knowledge of statistics

which comes under Topic 5: Assessment in statistics education

Other papers under session 5B ... 5B1: The statistics items in the Brazilian National Student Performance... 5B2: What do you know? Assessment beyond the traditional roles

(Tuesday 13th, 14:00-16:00)

Text analytic tools for the cognitve diagnosis of student writings

Presenter

• Tjaart Imbos (Maastricht University, Netherlands)

Co-author

• Ton Ambergen (Maastricht University, The Netherlands)

Abstract

Students can be stimulated to become active learners using a tool for active writing. In our university, we developed such a tool: POLARIS. Active writings of students about statistical concepts are valuable for the students and the teacher. In their writings, students show their understanding of statistical topics. The problem then is how to interpret and score the writings of students in relation to their proficiency in statistics. In this paper text analytic tools are used to cluster and score sample papers of students. Two approaches are compared: a statistics-based approach, Latent Semantic Analysis (LSA) and a linguistic-based approach, known as natural language processing (NLP). The key features of both approaches are discussed, as well as the usability, reliability and validity.

Other papers under session 5B ... 5B1: The statistics items in the Brazilian National Student Performance... 5B2: What do you know? Assessment beyond the traditional roles