Break it Down for Me: A Study in Automated Lyric Annotation

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2074–2080 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Break it Down for Me: A Study in Automated Lyric Annotation Lucas Sterckx*, Jason Naradowsky†, Bill Byrne‡, Thomas Demeester* and Chris Develder* * IDLab, Ghent University - imec firstname.lastname@ugent.be † Language Technology Lab, DTAL, University of Cambridge jrn39@cam.ac.uk ‡ Department of Engineering, University of Cambridge wjb31@cam.ac.uk Abstract Comprehending lyrics, as found in songs and poems, can pose a challenge to hu- man and machine readers alike. This moti- vates the need for systems that can under- stand the ambiguity and jargon found in such creative texts, and provide commen- tary to aid readers in reaching the correct interpretation. We introduce the task of automated lyric annotation (ALA). Like text simplifica- tion, a goal of ALA is to rephrase the original text in a more easily understand- able manner. However, in ALA the sys- tem must often include additional infor- mation to clarify niche terminology and abstract concepts. To stimulate research on this task, we release a large collec- tion of crowdsourced annotations for song lyrics. We analyze the performance of translation and retrieval models on this task, measuring performance with both au- tomated and human evaluation. We find that each model captures a unique type of information important to the task. 1 Introduction Song lyrics and poetry often make use of ambi- guity, symbolism, irony, and other stylistic ele- ments to evoke emotive responses. These charac- teristics sometimes make it challenging to inter- pret obscure lyrics, especially for readers or lis- teners who are unfamiliar with the genre. To ad- dress this problem, several online lyric databases have been created where users can explain, con- textualize, or discuss lyrics. Examples include MetroLyrics1 and Genius.com2. We refer to such 1http://www.metrolyrics.com 2http://genius.com How does it feel? To be without a home Like a complete unknown, Like a rolling stone The proverb “A rolling stone gathers no moss” refers to people who are always on the move, never putting down roots or accumulating responsibilities and cares. Figure 1: A lyric annotation for “Like A Rolling Stone” by Bob Dylan. commentary as a lyric annotation (Figure 1). In this work we introduce the task of auto- mated lyric annotation (ALA). Compared to many traditional NLP systems, which are trained on newswire or similar text, an automated system ca- pable of explaining abstract language, or finding alternative text expressions for slang (and other unknown terms) would exhibit a deeper under- standing of the nuances of language. As a result, research in this area may open the door to a va- riety of interesting use cases. In addition to pro- viding lyric annotations, such systems can lead to improved NLP analysis of informal text (blogs, so- cial media, novels and other literary works of fic- tion), better handling of genres with heavy use of jargon (scientific texts, product manuals), and in- creased robustness to textual variety in more tradi- tional NLP tasks …

Content truncated to 500 words