Komala and Kathora: A Novel Approach Towards Classification of Hindi Poetry

Literary compositions are very often analyzed using various constituent units like words, phrases, sentences, and paragraphs. Unlike the conventional research that focuses on the aforementioned constituent units, our task is a statistical effort carried out on the most fundamental unit of any literary composition called varna, or character, followed by automated classification using learning algorithms. This article is a case study on the Hindi adaptations of two significant literary pieces, namely, Jana-Gana-Mana and Vande-Mataram, and acknowledging that the two songs being studied belong to different classes based on their bhava, i.e., the inherent emotion of the poem. The present task is the first of its kind that uses the concept of komala and kat. hora varna to establish diversity between the two. The two-proportion Z-test is successfully applied to statistical data pertaining to the candidate songs, thereby reestablishing the theoretical assertions by investigating real pieces of literature. Taking the statistical verification as ground, a learning-based classification system is designed to yield the best accuracy of 85%, which further compliments the theory reestablished statistically.