Language Learnability Analysis of Hindi: A Comparison with Ideal and Constrained Learning Approaches

Native language acquisition is one of the initial processes undertaken by the human brain in the infant stage of life. The linguist community has always been interested in finding the method, which is adopted by the human brain to acquire the native language. Word segmentation in one of the most important tasks in acquiring the language. Statistical learning has been employed to be one of the earliest strategies that mimic the way an infant can adapt to segment a lot of different words. It is desired that the language learnability theories be universal in nature and work on most, if not all the languages. In the present work, we have analyzed the learnability of Hindi, the most popular Indian language, using ideal (universal) and constrained Bayesian learner models. We have analyzed the learnability of the language using unigram and bigram approaches by considering word, syllables, and phonemes as the smallest unit of the language. We demonstrate that Bayesian inference is indeed a viable cross-linguistic strategy and works well for Hindi also.