Building a Knowledge Base from Learning Objects
MetadataVis full innførsel
The goal of this thesis has been to investigate if it is possible to develop a knowledge structure,knowledge base, based on learning objects. In this connection a learning object is a digital unitwhich should, as a minimum, contain a picture and some text. Most likely a learning objectwould function as a container with anchors for video, animations and links to html-pages.For every learning object there exists a textual description. If we consider the learning object asan overhead presented in a classroom lecture, the text is a transcript of the explanation thelecturer would give, showing this overhead. This text should have a length of one half to a fullA4 page.As part of this thesis I have developed a search engine based on the vector model from the fieldof Information Retrieval.The idea is that all of the learning material in a course should exist as learning objects collectedin a repository. Based on the textual description tied to every learning object the contents of therepository will be represented as vectors in an n-dimensional vocabulary space.Using the search engine I will build an index for the collection of learning objects. Doing this,the engine will produce a similarity matrix that gives the similarity between all the learningobjects.Based on the similarity matrix I will visualize the learning objects in a three dimensionalknowledge structure.This visual structure will show all the learning objects and also how they are connected. Thestrength of the connection is determined from the similarity matrix. Two learning objects thathave a large similarity will be close in the 3D-graph.This presentation of all the learning objects in a course can be used in many different ways.When a lecturer is planning a course he or she can query the repository of learning objects. Thequery will be represented as a vector and placed in the knowledge structure. The lecturer islooking for the objects close to the query vector, but in addition the neighbouring objects wouldbe of interest.A student, solving a problem, can use the knowledge structure as a knowledge base.The advantage with this system is that it is built from statistic, calculating similarity betweenvectors. There is no need for manual treatment. This is also the weakness of the system, becausethere is no room for semantics.The experiments shows that the system has many interesting sides, but more research must beedone before it becomes a complete pedagogical system.