tailieunhanh - manning schuetze statisticalnlp phần 6

Để làm tốt nhiệm vụ này, việc xác định chính xác các mối quan hệ anaphoric là rất quan trọng để theo dõi những người tham gia. Bão Hugo bị phá hủy ngôi nhà Florida. Tại một chi phí ước tính của một tỷ đô la, thảm họa đã được nhất | 328 9 Markov Models State 1 2 3 T l Time t Figure Trellis algorithms. The trellis is a square array of states versus times. A node at Sị t can store information about state sequences which include xt i. The lines show the connections between nodes. Here we have a fully interconnected HMM where one can move from any state to any other at each step. The backward procedure It should be clear that we do not need to cache results working forward through time like this but rather that we could also work backward. The BACKWARD backward procedure computes backward variables which are the total PROCEDURE probability of seeing the rest of the observation sequence given that we were in state at time The real reason for introducing this less intuitive calculation though is because use of a combination of forward and backward probabilities is vital for solving the third problem of parameter The Three Fundamental Questions for HMMS 329 t t l Figure Trellis algorithms Closeup of the computation of forward probabilities at one node. The forward probability a J t 1 is calculated by summing the product of the probabilities on each incoming arc with the forward probability of the originating node. reestimation. Define backward variables fait Plot . . oT Xf 1 1P Then we can calculate backward variables working from right to left through the trellis as follows 330 9 Markov Models output lem ice_t cola Time t 1 2 3 4 cp t rp t P O1 . . Of-1 Pcp t P O1 Or Yep t y p t xt CP IP CP CP Scp t ỖIP t Ipcp t CP IP CP Wip t CP IP CP Xt CP IP CP CP P X Table Variable calculations for 0 lem ice_t cola . 1. Initialization 0 T 1 1 l i N 2. Induction N t X 1 1 t T 1 i N j i 3. Total JV P O p TT 1 Table shows the calculation of forward and backward variables and other