Modeling knowledge states in language learning

Show full item record



Permalink

http://urn.fi/URN:NBN:fi:hulib-202006172999
Title: Modeling knowledge states in language learning
Author: Vu, Anh-Duc
Other contributor: Helsingin yliopisto, Matemaattis-luonnontieteellinen tiedekunta
University of Helsinki, Faculty of Science
Helsingfors universitet, Matematisk-naturvetenskapliga fakulteten
Publisher: Helsingin yliopisto
Date: 2020
Language: eng
URI: http://urn.fi/URN:NBN:fi:hulib-202006172999
http://hdl.handle.net/10138/316599
Thesis level: master's thesis
Degree program: Tietojenkäsittelytieteen maisteriohjelma
Master's Programme in Computer Science
Magisterprogrammet i datavetenskap
Specialisation: Algoritmit
Algorithms
Algoritmer
Discipline: none
Abstract: Artificial intelligence (AI) is being increasingly applied in the field of intelligent tutoring systems (ITS). Knowledge space theory (KST) aims to model the main features of the process of learning new skills. Two basic components of ITS are the domain model and the student model. The student model provides an estimation of the state of the student’s knowledge or proficiency, based on the student’s performance on exercises. The domain model provides a model of relations between the concepts/skills in the domain. To learn the student model from data, some ITSs use the Bayesian Knowledge Tracing (BKT) algorithm, which is based on hidden Markov models (HMM). This thesis investigates the applicability of KST to constructing these models. The contribution of the thesis is twofold. Firstly, we learn the student model by a modified BKT algorithm, which models forgetting of skills (which the standard BKT model does not do). We build one BKT model for each concept. However, rather than treating a single question as a step in the HMM, we treat an entire practice session as one step, on which the student receives a score between 0 and 1, which we assume to be normally distributed. Secondly, we propose algorithms to learn the “surmise” graph—the prerequisite relation between concepts—from “mastery data,” estimated by the student model. The mastery data tells us the knowledge state of a student on a given concept. The learned graph is a representation of the knowledge domain. We use the student model to track the advancement of students, and use the domain model to propose the optimal study plan for students based on their current proficiency and targets of study.
Subject: Knowledge space theory
Bayesian knowledge tracing
GS algorithm
Bayesian network learning


Files in this item

Total number of downloads: Loading...

Files Size Format View
Msc_Thesis_2020.pdf 6.827Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record