Commonality and Diversity Analysis for Style Determination: Resource Creation and System Development

Stylometry has gained momentum within digital humanities, yet studies focusing on Hindi poetry remain scarce, especially in metrical compositions like dohā . This article presents a computational and statistical approach to analyzing stylistic commonality and diversity between poets, using dohā compositions as a case study. Addressing limitations in previous rhythm analyses that overlooked pronunciation, we introduce RISS (Rhythm Identification and Scoring System), which uses pronunciation-based mora counts. The RHISS , Rhyme Identification and Scoring System, an outcome of this study, employs a novel algorithm to identify rhyme in any pair of words in the Hindi language. These systems extract rhythmic and rhyming features, enabling detailed stylistic profiling. The analysis focuses on dohā by Tulsidas and Bihari, composed in the Awadhi and Brajbhasha dialects, respectively. A key contribution is the creation of speech-based resources for these under-resourced dialects. Comparative evaluation reveals that while both poets demonstrate similar mean rhyme scores, rhythm scores differ statistically significantly— 0.9824 for Tulsidas and 0.8832 for Bihari—indicating divergence in rhythmic structuring. Additionally, Tulsidas's work shows a higher proportion of dohās with optimal rhythm and a greater density of unique rhyming pairs, suggesting richer phonetic variation. These findings highlight both convergence and divergence in poetic style, grounded in linguistic and prosodic constraints.