Defence of dissertation in the field of Computer Science and Engineering, M.Sc. Lauri Lahti

Computer-Assisted Learning Based on Cumulative Vocabularies, Conceptual Networks and Wikipedia Linkage

M.Sc. Lauri Lahti, will defend the dissertation "Computer-Assisted Learning Based on Cumulative Vocabularies, Conceptual Networks and Wikipedia Linkage" on 10 April 2015 at 12 noon in Computer Science Building in lecture hall T2, Konemiehentie 2, Espoo. The dissertation proposes new methods and frameworks for computer-assisted learning based on developed and tested software prototypes that enable generation of guidance about useful learning paths in conceptual networks supported by various vocabularies and the link network of Wikipedia.

In a new doctoral dissertation Lauri Lahti proposes that to support learning expressive and adaptive knowledge structures can be created with conceptual networks  based on the structure of Wikipedia online encyclopedia. Already previous research has indicated structural similarities between Wikipedia, social networks and human brain networks. This refers to so called scale-free small-world properties that inherently emerge into collaboratively built Wikipedia, the world’s largest encyclopedia.

The doctoral dissertation proposes new methods for computer-assisted learning based on cumulative vocabularies, conceptual networks and linkage of Wikipedia articles. Relying on conceptual networks gathered from students and Wikipedia along the research diverse methods have been developed and compared for generation of automatic guidance about useful learning paths. Experimentally it appeared that when browsing in the link network of Wikipedia the students could recall linked words they selected to traverse about 95 percent better than alternative but non-selected on average 2.85 linked words at each step. Experimentally it appeared that by guiding the students to traverse the shortest link paths between concepts the students could recall the sentences corresponding to these links even 15 percent better than when exploring these sentences in a randomized order. When observing the recall in respect to pairs of consecutive links the guidance about the shortest link paths enabled even 26 percent better recall of the sentences than in a randomized order of the sentences.

Experimentally it was found that conceptual networks of students, common language and Wikipedia inherently emphasize different themes that should be addressed when developing learning methods. Along the research conceptual networks corresponding to eleven different language ability levels were created from Wikipedia to enable guidance about learning paths for diverse students in respect to vocabularies of five first school years and reaching at most an about 3000-word vocabulary that has been estimated to be possibly sufficient for a reasonable reading comprehension. Corresponding to vocabularies of about 250–3000 words a set of conceptual networks of about 1000–25000 links were created from Wikipedia and for learning paths relying on them some computational features were identified.

New methods to support learning proposed in the dissertation are contrasted with a broad review about measures of human learning process and representation of knowledge.

Dissertation press release (pdf)

Opponent: Associate Professor Piet Kommers, University of Twente, The Netherlands

Custos: Professor Jorma Tarhio, Aalto University School of Science, Department of Computer Science

Electronic dissertation: https://aaltodoc.aalto.fi/handle/123456789/15445

School of Science, electronic dissertations: https://aaltodoc.aalto.fi/handle/123456789/52