Abstract | In this paper we propose 1-best A*, 1-best iterative A*, k-best A* and k-best iterative Viterbi A* algorithms for sequential decoding. |
Abstract | In particular, we show that iterative Viterbi A* decoding can be several times or orders of magnitude faster than the state-of-the-art algorithm for tagging tasks with a large number of labels. |
Background | 3.1 l-Best Viterbi |
Background | The Viterbi algorithm is a classic dynamic programming based decoding algorithm. |
Introduction | Traditionally, the Viterbi algorithm ( Viterbi , 1967) is used. |
Introduction | Unfortunately, due to its 0(TL2) time complexity, where T is the input token size and L is the label size, the Viterbi decoding can become prohibitively slow when the label size is large (say, larger than 200). |
Introduction | The Viterbi algorithm cannot find the best sequences in tolerable |
Abstract | For tagging, learned constraints are directly used to constrain Viterbi decoding. |
Abstract | Dynamic programming techniques based on Markov assumptions, such as Viterbi decoding, cannot handle those ’nonlocal’ constraints as discussed above. |
Abstract | However, it is possible to constrain Viterbi |
A Class-based Model of Agreement | It can be learned from gold-segmented data, generally applies to languages with bound morphemes, and does not require a hand-compiled lexicon.3 Moreover, it has only four labels, so Viterbi decoding is very fast. |
Inference during Translation Decoding | Incremental Greedy Decoding Decoding with the CRF—based tagger model in this setting requires some slight modifications to the Viterbi algorithm. |
Inference during Translation Decoding | This forces the Viterbi path to go through If. |