Date of Degree


Document Type


Degree Name



Computer Science


Danny Kopec

Subject Categories

Computer Sciences | Instructional Media Design


big data, ESL, graphical organizers, HCI, human-computer interaction, intelligent tutoring systems


Reading comprehension is predominately measured through multiple choice examinations. Yet, as we will discuss in this thesis, such exams are often criticized for their inaccuracies. With the advent of "big data" and the rise of ITS (Intelligent Tutoring Systems), increasing focus will be placed on finding dynamic, automated ways of measuring students' aptitude and progress.

This work takes the first step towards automated learner classification based on the application of graphic organizers. We address the following specific problem experimentally: How effectively can we measure task comprehension via human translation of written text into a visual representation on a computer? Can an algorithm employ data from user interface (UI) interaction during the problem solving process, to classify the user's abilities? Specifically, from the data we show machine learning predictions of what a human expert would say about the:

1. integrity of the visual representation produced;

2. level of logical problem solving strategy the user applies to the exercise;

3. level of effort the user gives to the exercise.

The core of the experiment is a software system that allows a human subject to read a preselected text and then "draw" a diagram by manipulating icons on a grid-canvas using standard transforms.