|
|
Quantitative Approaches to Versification
Plecháč, Petr ; Barry, P. ; Skulacheva, T. ; Bermúdez-Sabel, H. ; Kolár, Robert
This volume presents a wide range of quantitative approaches to versification. It comprises various methodological perspectives ranging from simple descriptive statistics to advanced machine learning methods (such as support vector machines, random forests or neural networks) as well as material covering a large span of time and languages: from very ancient versifications (Sumerian, Akkadian, Hittie, Ancient Greek), through medieval (Old English, Old Icelandic, Old Saxon) and Renaissance verse to modern experiments (free verse, concrete poetry), from English and Russian through Spanish and German to Portuguese and Catalan. Not only written, but also spoken poetry has been analyzed. The book covers multiple topics. What they all share in common is that versification is being studied in the context of other linguistic phenomena that may affect or determine it. Analyses of large corpora go hand in hand with comparative approaches. It is shown that quantitative approaches can be used for the purpose of authorship attribution, to build reasonable typologies as well as to understand why certain forms play such a dominant roles in our cultural tradition(s).
|
|
|
Assessing the reliability of stress as a feature of authorship attribution in syllabic and accentual syllabic verse
Plecháč, Petr ; Birnbaum, D. J.
This work builds on a recent study by one of the authors, which shows that statistics about versification may be used as a feature in the process of authorship attribution. One such statistic is what we have called the stress profile of a poem, a vector consisting of frequencies of stressed syllables at particular metrical positions. Our initial hypothesis was that because syllabic versification (SV) regulates by definition the number of syllables in a line but not the distribution of stresses, it allows authors to individualize their rhythmical style much more than accentual syllabic versification (ASV), where the distribution of stresses is primarily determined by meter. For that reason, we expected the stress profile to be a more reliable indicator of authorship in Spanish SV than in Czech or German ASV. This hypothesis, however, was not supported by our analysis. For most of our samples, German ASV had lower accuracy than Spanish, which we had predicted, but, contrary to our expectations, the accuracy for Czech ASV and Spanish SV were more or less the same. This result led us to hypothesize further that the traditional labels SV and ASV were misleading and we sought to measure the tonic entropy of our data. In this case, Spanish SV, as expected, was found to be the least tonically regular, while there was a significant difference between the two ASV systems: the values for Czech were even closer to Spanish than to the low-scoring German system. This explains why our initial grouping of Czech and German together into a single ASV category was insufficiently nuanced.
|
host ::
RSS.