Rhyme is a relevant structural element in many poetic forms. Besides its aesthetic and musical function, rhyme words hold a privileged position to convey meaning, and the relation created between sound and sense must not be overlooked (Houston, 2016). Considering the use of rhyme as a mnemonic device (e.g. Rubin, 1997) and its role in vocabulary retrieval (e.g. Rapp & Samuel, 2002), the evaluation of the semantic properties of rhyme words becomes crucial for a holistic understanding of the poetic composition. We implemented natural language processing methods on a corpus of Spanish sonnets (15th–17th centuries) to examine the semantics of rhyme words in terms of sentiment analysis, i.e. positive or negative connotation (Augustyniak et al., 2016). The analysed corpus comprises the Corpus of Spanish Golden-Age Sonnets (Navarro-Colorado et al., 2016) and the corresponding period of the Diachronic Spanish Sonnet Corpus (Ruiz Fabo et al., 2020). Sentiment analysis on rhyme words was performed to detect stylistic patterns within the work of an author and across poets. A second experiment assessed whether rhyme words were used to reinforce a sentiment or to create a contrast by analysing the rest of the poem. We implemented a lexicon-based approach, updating ad hoc the lemmas available in the Spanish section of the ML-SentiCon lexicon (Cruz et al., 2014). This allowed us to account for the specificities of the historical variant under analysis and particular uses of the poetic language. Our study sheds light on the semantic properties of rhyme in the Spanish Golden Age sonnet, presenting a methodology easily adapted to other poetic forms, languages and periods, which can contribute to the ongoing debate around the definition of this rhetorical device concerning its semantic dimension.