Main Article Content
Language processing make easier path to various speech recognition system to enhance the precision of various Language interaction systems. In this paper we project the work done on updating and improvement of the language model component of a continuous speech recognition system. This experiment proposes a recurrent neural network based language model for Tamil speech recognition system, which aims at reduction of perplexity, word error rate and generating more meaningful text. The RNN language model operates as a predictive model for the next word given in the previous schema. In our experiments, we show that the recurrent neural network language model outperforms the n-gram model and feed forward language model on various Tamil language datasets.