MELODY ANALYSIS AND CONTINUATION
This report presents a novel approach to automatic music simulation by extraction of musical data and subsequent comparison on this data for continuation. the research and development of the use of natural language processing for continuation in a prototype named ” music continuation” for continuation is documented. background research has been reported on systems that produce musical continuation, with particular attention to the solutions developed on related problems by the MOSART group and francois pachet. Research is also discussed on pattern detection techniques with particular focus on n-gram pattern analysis in natural language processing.
the specification of the n-gram technique is documented for the capture of musical patterns from MIDI files and inputted note information. the subsequent data formatting of the n-gram data into a suitable frequency representation for continuation and why it is done is also documented. Algorithms are presented for the generation of continuation notes and subsequent continuation using these notes. The report also suggests future development directions of the system, particularly extentions to the number of musical pattern types encoded and optimisation of their use, stylistic play, polyphonic continuation and other facets that time was not avaialable to implement or research. a report of the evaluation process and it findind are included, as are the conclusions of the suitability of the n-gram corpus technique