Machine Learning Algorithm for the Scansion of Old English and Old Saxon Poetry

Abstract

Several scholars designed tools to perform the automatic scansion of poetry in many different languages; however, none of these tools deals with Old English and Old Saxon meter. Therefore, the goal of this thesis was to develop and implement a Bidirectional Long Short Term Memory model to perform the automatic scansion of Old English and Old Saxon poems. Since this model is a supervised machine learning model; the first step of this thesis was to create a metrically annotated corpus of Old English. The 6000 verses of Heliand compose this corpus, which was annotated following Suzuki's annotation of Heliand from "The Metre of Old Saxon Poetry: The Remarking of Alliterative Tradition" (2004). The second step of this thesis consisted in training a Bidirectional Long Short Term Memory, in order to learn to identify the metrical patterns of the verses and to assign to each verse its correct metrical type.