Literary and Linguistic Computing Advance Access originally published online on March 3, 2006
Literary and Linguistic Computing 2006 21(Supplement 1):127-141; doi:10.1093/llc/fql011
| ||||||||||||||||||||||||||||||||||||||||||||||||
Virtual Verse Analysis: Analysing Patterns in Poetry
Department of English, University of Toronto, Canada
Correspondence: Marc R. Plamondon, 130 St George Street, 7th floor, John P. Robarts Research Library, University of Toronto, Toronto, Ontario, M5S 1A5. E-mail: mplamond{at}chass.utoronto.ca
This article discusses the problem of computer identification of some basic patterns in poetry: rhythm and rhyme. The program AnalysePoems uses the Representative Poetry Online (http://rpo.library.utoronto.ca) corpus of poems to identify these patterns. The goal of the rhythm identification is not to produce a perfect metrical scansion of a poem, but to identify the dominant metre with a reasonable degree of confidence. Twelve example poems are used to show that the program is successful in its ability to identify the dominant metre, if one exists, even when some of the words of the poem are unknown to the computer. The computer is usually able to identify the number of syllables and dominant accent of these previously unknown words and then use this new data in its analysis of more poems. Similarly, an analysis of a poem's rhyme scheme allows the computer to identify rhyming pairs it had not previously encountered and use this new information to analyse subsequent rhyme schemes. The analysis of basic patterns lays the groundwork for the analysis of more complex, less obvious patterns. The changes in rhythmic confidence values from first to final analysis suggest a possible measure of the complexity and beauty of a poem's rhythms.