Autonomous Semantic Structuring of Lecture Topics Synthesis of Knowledge Models Robin Nicolay, Nikolaj Troels Graf von Malotky, Tanja Auge and Alke Martens University of Rostock, Institute of Computer Science, Chair of Practical Informatics, 18051 Rostock, Germany robin.nicolay@uni-rostock.de, nikolaj.graf von malotky@uni-rostock.de Keywords: Latent Dirichlet Allocation, Topic Models, Mental Models, Knowlege Management, Force-directed Algo- rithms. Abstract: Students attending lectures in universities suffer from a weak structural awareness on lecture content. Ac- cording to learning theories, structural awareness is a relevant factor to association and comprehension of new learning inputs. We synthesize semantic structures from non annotated lecture slides using Topic Modeling algorithms to identify relevant terms and relate them in force-directed graphs. The synthesized graphs provide a structural overview on the topic distribution and relations of non annotated sequential lecture slides. 1 INTRODUCTION University teachers, aka lecturers, use lectures to teach facts and concepts to students. From the per- spective of the teacher, in the preparation phase the relevant information of a lecture is analyzed, reduced and sequenced into learning units and taught in a way that is comparable to leading a path through a con- strained area of knowledge. During a study a stu- dent learns from many different lectures or knowledge units. Some of these units reference each other, some are partially overlapping or describe common topics from different perspectives. An ideal situation would be given, if the students identifies and understands the key concepts plus their relations and their hierarchy or topology. However, reality often leads to different situations: Even if the teacher usually is preparing a lecture via sequencing the important facts of an over- all topic in a logical order, in quite a lot cases, lec- tures are evolutionary grown over time. This results in a situation, where the key concepts of a lecture are somewhat hidden in the text. From the perspective of the teacher, this is not so bad, as the overall pic- ture shall not be influenced. However, in exams, we were able to observe that students often miss impor- tant items or misinterpret items. The situation grows even worse, when students were asked to detect rela- tions between different topics, i.e. different lectures in different semesters. This seems to be a cognitive step which is not directly supported in current lecture formats. On the teacher’s side, an interconnection of the topics or areas of study is seldom taking place. This results in a situation where students are usually missing the overall picture. To develop a solution, i.e. to develop support mechanisms and tools to support the student’s knowl- edge construction in lectures, we have to take a look at learning psychology. Learning theories such as cognitivism describe the inner effects of processing lecture information inputs by using cognitive models (lernpsychologie.net, 2016). To understand, how pro- cesses of knowledge construction could potentially take place, we made several investigations together with our students. Our work started a while ago, when we developed a tool for extracting knowledge from lectures with the goal to support student’s annotation of lectures and for giving support for learning (Nicolay et al., 2015). To support our claim that understanding of the main concepts and their interrelation is the key to under- standing the lecture in a first step and to understand- ing the overall picture in a study direction as a sec- ond step, we investigated student’s intuitive method of knowledge construction. Our main study took place last semester, where we made a structured investigation with a first grade master course with 20 students from different depart- ments. They were organized in teams of two to four (mixed male and female). They got a free choice of material (either digital or not), and free choice of most important topics. Their task was to identify areas of knowledge and relations in self designed knowledge Nicolay, R., Malotky, N., Auge, T. and Martens, A. Autonomous Semantic Structuring of Lecture Topics - Synthesis of Knowledge Models. DOI: 10.5220/0006367903490355 In Proceedings of the 9th International Conference on Computer Supported Education (CSEDU 2017) - Volume 2, pages 349-355 ISBN: 978-989-758-240-0 Copyright © 2017 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved 349