Are Proverbs the New Pythian Oracles? Exploring Sentiment in Greek Sayings

TL;DR

This study applies NLP, including in-context learning with LLMs, to analyze sentiment in Greek proverbs and extend coverage to local dialects. By building a standard-plus-dialect dataset, mapping proverbs geographically, and evaluating multiple prompting strategies, the authors show that LLMs can approximate proverb sentiment with meaningful accuracy, though sentiment interpretation remains inherently subjective. The findings reveal a regional tilt toward negative sentiment and highlight data gaps in urban areas, underscoring the value and limitations of NLP for paremiology. Overall, the work lays a foundation for dialect-aware sentiment mapping of proverbs and demonstrates practical methods for digital humanities research in low-resource languages.

Abstract

Proverbs are among the most fascinating linguistic phenomena that transcend cultural and linguistic boundaries. Yet, much of the global landscape of proverbs remains underexplored, as many cultures preserve their traditional wisdom within their own communities due to the oral tradition of the phenomenon. Taking advantage of the current advances in Natural Language Processing (NLP), we focus on Greek proverbs, analyzing their sentiment. Departing from an annotated dataset of Greek proverbs, we expand it to include local dialects, effectively mapping the annotated sentiment. We present (1) a way to exploit LLMs in order to perform sentiment classification of proverbs, (2) a map of Greece that provides an overview of the distribution of sentiment, (3) a combinatory analysis in terms of the geographic position, dialect, and topic of proverbs. Our findings show that LLMs can provide us with an accurate enough picture of the sentiment of proverbs, especially when approached as a non-conventional sentiment polarity task. Moreover, in most areas of Greece negative sentiment is more prevalent.

Figures (6)

• Figure 1: Top 20 topics according to the labels found in the gnomikologikon. The topics are translated from Greek to English.
• Figure 2: Pearson correlation heatmap in the sentiment annotations. From a1 to a13 are the annotators.
• Figure 3: Predicted positive and negative sentiment maps of Greece. In both cases, ambiguous instances are aggregated into the positive and negative classes. Interactive versions of the maps, including proverb counts and the option to view individual proverbs, will be released upon acceptance.
• Figure 4: Spearman correlation heatmap in the sentiment annotations.
• Figure 5: Cohens correlation heatmap in the sentiment annotations.