A seminatural approach towards restoration, classification, and pattern extraction from heritage Hindi poetry: A probabilistic case study with Doha chanda

Indian literary heritage is vast and is of great importance; to explore it, one has to devote them in studying dialects (Khariboli, Haryanavi, Brajbhasha, Awadhi, Bhojpuri, Marwari, etc.) especially when old Hindi is under the lens of observation. Chanda are poetic compositions that have well-defined structures. Doha is a kind of chanda which, in our work, has been explored using Kabir's compositions as a case study. Kabir represents the cult of poets who relied on oral means for the propagation and consumption of poetry. The poems were communicated to the later generation through simple acts of recitation and hearing, leading to obvious mutation and multiple versions of the same compositions when documented later and hence needs restoration (at least metrically). Using the knowledge from the state-of-art models metadata generator, Text2Matra, and RPaGen and extending beyond them, this article is first of its kind to present Kabir's doha within the scope of restoration, metrical computation, and statistics. Starting with the restoration process, the proposed algorithms generate data that are subjected to suitable statistical models. These models highlight the trends in the dataset, which is helpful in realizing abstract patterns inside the textual data.