Automatic Analysis of Rhythmic Poetry with Applications to Generation and Translation

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 524–533, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Automatic Analysis of Rhythmic Poetry with Applications to Generation and Translation Erica Greene Haverford College 370 Lancaster Ave. Haverford, PA 19041 ericagreene@gmail.com Tugba Bodrumlu Dept. of Computer Science Univ. of Southern California Los Angeles, CA 90089 bodrumlu@cs.usc.edu Kevin Knight Information Sciences Institute Univ. of Southern California 4676 Admiralty Way Marina del Rey, CA 90292 knight@isi.edu Abstract We employ statistical methods to analyze, generate, and translate rhythmic poetry. We first apply unsupervised learning to reveal word-stress patterns in a corpus of raw poetry. We then use these word-stress patterns, in ad- dition to rhyme and discourse models, to gen- erate English love poetry. Finally, we trans- late Italian poetry into English, choosing tar- get realizations that conform to desired rhyth- mic patterns. 1 Introduction When it comes to generating creative language (po- ems, stories, jokes, etc), people have massive advan- tages over machines: • people can construct grammatical, sensible ut- terances, • people have a wide range of topics to talk about, and • people experience joy and heart-break. On the other hand, machines have some minor ad- vantages: • a machine can easily come up with a five- syllable word that starts with p and rhymes with early, and • a machine can analyze very large online text repositories of human works and maintain these in memory. In this paper we concentrate on statistical methods applied to the analysis, generation, and translation of poetry. By analysis, we mean extracting patterns from existing online poetry corpora. We use these patterns to generate new poems and translate exist- ing poems. When translating, we render target text in a rhythmic scheme determined by the user. Poetry generation has received research attention in the past (Manurung et al., 2000; Gervas, 2001; Diaz-Agudo et al., 2002; Manurung, 2003; Wong and Chun, 2008; Tosa et al., 2008; Jiang and Zhou, 2008; Netzer et al., 2009), including the use of statistical methods, although there is a long way to go. One difficulty has been the evaluation of machine-generated poetry—this continues to be a difficulty in the present paper. Less research effort has been spent on poetry analysis and poetry trans- lation, which we tackle here. 2 Terms Meter refers to the rhythmic beat of poetic text when read aloud. Iambic is a common meter that sounds like da-DUM da-DUM da-DUM, etc. Each da- DUM is called a foot. Anapest meter sounds like da-da-DUM da-da-DUM da-da-DUM, etc. Trimeter refers to a line with three feet, pentame- ter to a line with five feet, etc. Examples include: • a VE-ry NAS-ty CUT (iambic trimeter) • shall I com-PARE thee TO a SUM-mer’s DAY? (iambic pentameter) • twas the NIGHT before CHRIST-mas and ALL through the HOUSE (anapest tetrame- ter) Classical English sonnets are poems most often composed of 14 lines of iambic pentameter. 524 3 Analysis We focus on English rhythmic poetry. We define …

Content truncated to 500 words