M.Sc. Jukka Toivanen will defend his doctoral thesis Methods and Models in Linguistic and Musical Computational Creativity in the University of Helsinki Main Building, Auditorium XIV (Unioninkatu 34, 3rd floor) on Friday the 3rd of June, 2016 at 12 o'clock. His opponent is Doctor Hisar Maruli Manurung (Google Japan Inc., Japan) and custos Professor Hannu Toivonen (University of Helsinki). The defence will be held in English.

Methods and Models in Linguistic and Musical Computational Creativity

Computational creativity is an area of artificial intelligence that develops algorithms and simulations of creative phenomena, as well as tools for performing creative tasks. In this thesis, we present various computational methods and models of linguistic and musical creativity. The emphasis is on developing methods that are maximally unsupervised, i.e. methods that require a minimal amount of hand-crafted linguistic, world, or domain knowledge.

This thesis consists of an introductory part and five original research articles. The introductory part outlines computational creativity as a research field and discusses some of the philosophical foundations underlying the current work. The research articles present specific methods and algorithms for automatic composition of poetry and songs. The first article proposes a corpus-based poetry generation method that relies on statistical language modelling and morphological analysis and synthesis. In the second article, we expand that basic model with constraint programming techniques to handle more aspects of the poetic structure and style. The third article presents a method for mining document-specific word associations and proposes using them in poetry generation to produce poems based, for instance, on a specific news story. The fourth article presents a song composition system that utilises constraint programming to produce songs with matching lyrics and music in a transformational way, i.e. it is able to modify its own search space and preferences. Transformationality of the system is achieved with a metalevel component that can modify the system's internal constraints leading into new conceptual spaces. Finally, the fifth article discusses possibilities of combining personal biosignal measurements, especially electroencephalography, with techniques of computational creativity and presents an art installation called Brain Poetry based on these ideas.

The current work relies heavily on the use of unsupervised data mining techniques to automatically build models of specific creative domains such as poetry. The proposed methods and models are flexible and they are to a large extent independent of language and style. Thus, they provide a general framework for computational or synthetic creativity in linguistic and musical domains that can be easily expanded in many ways. Applications of this work include pedagogical tools, computer games, and artistic results.

Availability of the dissertation

An electronic version of the doctoral dissertation is available on the e-thesis site of the University of Helsinki at http://urn.fi/URN:ISBN:978-951-51-2204-9.

Printed copies will be available on request from Jukka Toivanen: jukka.toivanen@helsinki.fi.