Supervised Machine Learning for Hybrid Meter

Abstract

Following classical antiquity, European poetic meter was complicated by traditions negotiating between the prosodic stress of vernacular dialects and a classical system based on syllable length. Middle High German (MHG) epic poetry found a solution in a hybrid qualitative and quantitative meter. We develop a CRF model to predict the metrical values of syllables in MHG epic verse, achieving an Fscore of .894 on 10-fold cross-validated development data (outperforming several baselines) and .904 on held-out testing data. The method used in this paper presents itself as a viable option for other literary traditions, and as a tool for subsequent genre or author analysis.

A Metrical Analysis of Medieval German Poetry Using Supervised Learning

“…33 "in [these books] he began to search, " developing out of classical antiquity, makes it difficult to scan poetry using these methodologies, and thus supervised learning presents itself as an attractive method. After initial results of this project were published in Estes and Hench (2016), similar studies were undertaken for English in Agirrezabal et al (2016), Spanish in Navarro (2017), and Portuguese in Mittmann (2016) with the results here serving as the benchmark.…”

“…In a sense, the bare model is a model best suited to predicting stress. 39 Results for the model without constraints are an F-score of 0.894 on crossvalidated data and 0.904 on held-out data (Estes and Hench, 2016). 40 Generally, this rules out the stumpf (blunt) cadence, which carries only three stresses.…”

Metrical Tagging in the Wild: Building and Annotating Poetry Corpora with Rhythmic Features

“…Earlier work (Nenkova et al, 2007) already found strong evidence that part-of-speech tags, accentratio 22 and local context provide good signals for the prediction of word stress. Subsequently, models like MLP (Agirrezabal et al, 2016), CRFs and LSTMs (Estes and Hench, 2016;Agirrezabal et al, 2019) and transformer models (Talman et al, 2019) have notably improved the performance to predict the prosodic stress of words and syllables. Unfortunately, most of this work only evaluates model accuracy on syllable or word level, with the exception of Agirrezabal et al (2019).…”

“…While the speech processing community explores end-to-end methods to detect and control the overall personal and emotional aspects of speech, including fine-grained features like pitch, tone, speech rate, cadence, and accent (Valle et al, 2020), applied linguists and digital humanists still rely on rule-based tools (Plecháč, 2020;Anttila and Heuser, 2016;Kraxenberger and Menninghaus, 2016), some with limited generality (Navarro-Colorado, 2018;Navarro et al, 2016), or without proper evaluation (Bobenhausen, 2011). Other approaches to computational prosody make use of lexical resources with stress annotation, such as the CMU dictionary (Hopkins and Kiela, 2017;Ghazvininejad et al, 2016), are based on words in prose rather than syllables in poetry (Talman et al, 2019;Nenkova et al, 2007), are in need of an aligned audio signal (Rosenberg, 2010;Rösiger and Riester, 2015), or only model narrow domains such as iambic pentameter (Greene et al, 2010;Hopkins and Kiela, 2017;Lau et al, 2018) or Middle High German (Estes and Hench, 2016).…”

Phonological Soundscapes in Medieval Poetry

“…For this implementation, onsets were deemed illegal if they appeared in less than .02% of all onsets in the corpus. This threshold helps reduce the acceptance of onsets in foreign words.13 SeeEstes and Hench (2016) for a computational approach to MHG phonology and meter.14 Excepted are several end syllables in divided falls such as '-er', '-el', and 'ez'(Domanowski et al, 2009).15 Wolfram von Eschenbach (1994, 17, l. 5-21) "This is a peculiar arrangement. "16 This last change is important for our analyses in that it changes an open syllable ('lî') to a closed syllable ('lîc').…”