The University of Sheffield
Department of Computer Science

Mikhail Molotkov Undergraduate Dissertation 2016/17

Musification of Book

Supervised by J.Barker

Abstract

Everyone is familiar with visualisation tools for various data. However, what if we were able to represent data in musical form?

Present project aims to find a way to visualize books via music form, rather than actual visualisation, such as: histograms, word frequency diagrams, word clouds and etc. With help of modern techniques for text analysis and music composition by neural networks, the ultimate goal of the research shall be reached - design and develop the system, that would be capable to map features, extracted from the text, onto music parameters in a way, so it would be intuitively easy for untrained person to recognise text patterns that are presented in the musical form.

This paper would make it way through all stages: starting with motivation and initial idea, that has been modified to suit this project, discussing available and most appropriate techniques for text and music analysis: lexical density, sentiment and readability scores for text analysis and key identification and complexity score computation for music analysis respectively; and then available choices for neural networks: RNN and LSTM networks - that would be used for music composition, based on classical music. Straight after that, most suitable approaches would be chosen to make this project feasible and realistic for the given time frame and moreover, requirements would be identified, that should be met and fulfilled in order for the project to be built on time.

Once theoretical background is identified, paper would cover actual design and implementation of the system, where flowcharts would be presented for each module: text analysis, neural network, music analysis, mapping and music composition modules - along with sufficient description about how each module operates and relates to one another. Then evaluation and results would be covered along with proposed improvements. Thereafter, final conclusion would be drawn to finish up the research paper.