Build a Laplace-smoothed n-gram language model from tokenized training sentences and score test sequences.
