Problem 2: Trigram model We can increase the context sensitivity of our model without making it too much more complicated by moving to a trigram model. Using the chain rule again, we have: P(live, in, their, house, today) = P(live) × P(in|live) × P(their|in, live) × P(house|their, in) × P(today house, their) And as before, we can simplify this to: P(their, house, today|in, live) = P(their|in, live) × P(house their, in) × CT (live in their) CB (live in) Х CT (in their house) CB (in their) P(today house, their) CT (their house today) CB (their house) (1) (2) where CT and CB are trigram and bigram frequencies. For this problem, implement and evaluate a trigram-based confusable solver. Δ

Fig: 1