The idea of project was to automatically complete the sentences by learning a previous data-set. We used a hybrid approach to build a word predictor based on statistical analysis and then drilling down the result of this model by syntactic analysis for better results. In this, n-gram model is used as base for statistical analysis. The model offers, via a nonlinear recursive procedure, a computational and space efficient solution to the problem of estimating probabilities from sparse data. This analyses output a pretty large set of results with their corresponding probabilities. The probabilities are then used in conjunction with the Viterbi algorithm to find the most probable sequence of part-of-speech tags for a given sentence. Viterbi algorithm states that the probability of transitioning into a state representing POS tag t(i) depends upon previous two POS tags, tag(i-1) and tag(i-2), So it helps to make word prediction precise based on previous words and drill results to a very specific and precise range.