Finding Fingerprints in the Fabric of Verse: Unearthing Style in Old English Poetry Using Machine Learning
Abstract
This study hypothesizes that poetic styles exist in Old English literature and that these styles, whether they pertain to an individual or a school of poets, can be identified and categorized by examining verse syntax, particularly the usage of auxiliary verbs and verbals. Further, this thesis attempts to determine what common features exist in the handful of poems currently accepted as having been composed by the poet Cynewulf and what other works share this style.Scholars have previously undertaken such studies of style by investigating word-order (e.g., Bliss 1980) and by meticulously cataloging and analyzing data about auxiliaries, verbals, their types, and the environment in which they occur (e.g., Donoghue 1987). By and large they have found that classifications, and therefore styles, of Old English poems are not clear-cut.
This thesis builds upon such previous work and analyzes the same data by adopting an unsupervised machine learning approach which uses computer algorithms to cluster 19 poems into groups based on their syntactic features, and to determine if these groups represent styles based on the shared characteristics of their constituent poems.
This analysis did not find distinctive styles in the clusters that emerged from the larger corpus that was studied, indicating that Old English poetry was not monolithic but consisted of a range of styles throughout its history. In the narrower context of Cynewulfian authorship, the results were more promising and suggest that an additional poem can be considered for inclusion in his canon.
Terms of Use
This article is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAACitable link to this page
http://nrs.harvard.edu/urn-3:HUL.InstRepos:37736779
Collections
- DCE Theses and Dissertations [1189]
Contact administrator regarding this item (to report mistakes or request changes)