AriEmozione 2.0: Identifying Emotions in Opera Verses and Arias

Abstract

We present the task of identifying the emotions conveyed by the lyrics of Italian opera arias. We shape the task as a multi-class supervised problem, considering the six emotions from Parrot's tree: love, joy, admiration, anger, sadness, and fear. We manually annotated an opera corpus with 2.5k instances at the verse level and experimented with different classification models and representations to identify the expressed emotions. Our best-performing models consider character 3-gram representations and reach relatively low levels of macro-averaged F 1 . Such performance reflects the difficulty of the task at hand, partially caused by the size and nature of the corpus: relatively short verses written in 18th-century Italian. Building on what we learned from the verse-level setting, we adopt a higher granularity and increase the size of the corpus. First, we switch from verses to arias in order to have longer and more expressive texts. Second, we construct a new corpus with 40k arias (∼ 90k verses). This new dataset contains silver data, annotated by self-learning on the basis of an ensemble of binary classifiers.We then experiment with more sophisticated representations, by learning an embedding space and using it to train new models for the identification of emotions at the aria level, obtaining a significant performance boost.

The Sentiment of Latin Poetry. Annotation and Automatic Analysis of the Odes of Horace

“…Ancient Greek, is the one on Aeschylus, while the annotation of the Iliad is made of texts translated into Modern Greek. Additional interesting annotated resources are those containing other textual genres based on versification, that is songs (Çano and Morisio 2017;Apoorva and Radhika 2018) and operas: as for the latter, AriEmozione 2.0 is made of lines taken from Italian 17th-and 18th-century operas (Zhang et al 2022) annotated with one out of 6 emotions (love, joy, admiration, anger, sadness, fear).…”

“…Each tweet is annotated as neutral or as one, or more, of eleven emotions. The original categories were mapped onto our four sentiment classes, leaving out ambiguous emotions; r AriEmozione 2.0: lines taken from a set of Italian opera texts annotated with one out of six emotions (Zhang et al 2022). We mapped the original emotions to either the positive class or the negative class.…”