A reputation system assists people selecting whom to trust, encourages trustworthy action, and discourages participation of unskilled or dishonest. The “all good reputation” problem is common in current reputation systems, especially in e-commerce domain, making it difficult for buyers to choose credible sellers. Observing high growth of online data in Hindi language, in this paper, we propose a reputation system in this language. The functions of this system include (1) review mining for different criteria of online transactions, (2) calculation of reputation rating using Bayesian method, (3) calculation of reputation weight using typed dependency relation representation and Latent Dirichlet Allocation topic modeling technique for each criteria from user reviews, and (4) ranking sellers based on computed reputation score. Extensive simulations conducted on eBay dataset and TripAdvisor dataset show its effectiveness in solving “all good reputation” problem. So far, as our knowledge is concerned, this is the first work in Hindi language on reputation system.