Style Detection for Free Verse Poetry from Text and Speech

Proceedings of the 27th International Conference on Computational Linguistics, pages 1929–1940 Santa Fe, New Mexico, USA, August 20-26, 2018. 1929 Style Detection for Free Verse Poetry from Text and Speech Timo Baumann Language Technologies Institute Carnegie Mellon University Pittsburgh, USA tbaumann@cs.cmu.edu Hussein Hussein and Burkhard Meyer-Sickendiek Department of Literary Studies Free University of Berlin Berlin, Germany {hussein,bumesi}@zedat.fu-berlin.de Abstract Modern and post-modern free verse poems feature a large and complex variety in their poetic prosodies that falls along a continuum from a more fluent to a more disfluent and choppy style. As the poets of modernism overcame rhyme and meter, they oriented themselves in these two opposing directions, creating a free verse spectrum that calls for new analyses of prosodic forms. We present a method, grounded in philological analysis and current research on cognitive (dis)fluency, for automatically analyzing this spectrum. We define and relate six classes of poetic styles (ranging from parlando to lettristic decomposition) by their gradual differentiation. Based on this discussion, we present a model for automatic prosodic classification of spoken free verse poetry that uses deep hierarchical attention networks to integrate the source text and audio and predict the assigned class. We evaluate our model on a large corpus of German author-read post- modern poetry and find that classes can reliably be differentiated, reaching a weighted f-measure of 0.73, when combining textual and phonetic evidence. In our further analyses, we validate the model’s decision-making process, the philologically hypothesized continuum of fluency and investigate the relative importance of various features. 1 Introduction One of the most important explanations for modern art is the theory of aesthetic pleasure, which claims that the fluency of cognitive processing is the cause for the positive effect of aesthetic experience (Topolinski and Strack, 2009). Similarly, cognitive research on fluency showed that people rate stimuli that are processed more easily higher (Belke et al., 2010). On the other hand, many modern artists like Picasso or Sch¨onberg complicated the processability of their works using processes of abstraction in order to prevent such automated or fluid forms of art comprehensibility. Regarding this development, Bullot and Reber introduced the term disfluency as an artistic strategy to bring more analytical forms of art experience to the fore (Bullot and Reber, 2013). Other researchers suggested that disfluency prompts people to process information more carefully, deeply, and on a higher level of abstraction (Smith and Smith, 2006). We present a computational analysis that tests these two trends in the current discussion on art experience by focusing on the prosodic feature of (dis)fluency in modern and post-modern poetry. We assume that the rhythmical quality of modern poetry is “not properly measurable by the rules of traditional verse” (Wesling, 1996), since modern poetry often rejects the traditional metrical verse. Most modern and post-modern poetry uses rhythmical features beyond the metrical forms, preferring the imitation of naturalness of everyday language and its very fluent speech prosody. At the same time, modern poetry uses disfluent processes to introduce the kinds of obstructions to ease consumption …

Content truncated to 500 words