## STATISTICAL MACHINE TRANSLATION DECODER BASED ON PHRASE

### BibTeX

@MISC{Watanabe_statisticalmachine,

author = {Taro Watanabe and Eiichiro Sumita},

title = {STATISTICAL MACHINE TRANSLATION DECODER BASED ON PHRASE},

year = {}

}

### OpenURL

### Abstract

This paper describes a decoding algorithm for statistical machine translation based on phrases. In the past, the solution to the decoding problem were inspired from that of speech recognizers, translating each input word into one or more output words generating in left-to-right direction. The algorithm presented here iteratively constructs phrases or chunks of cepts until all the input words are consumed. This behavior resulted in computational complexity higher than those with left-to-right constraints, though the translation accuracy is better from the Japanese-to-English translation experiments. 1.

### Citations

1212 | The mathematics of statistical machine translation: Parameter estimation
- Brown, Pietra, et al.
- 1993
(Show Context)
Citation Context ... the translation is the maximum likelihood solution of both of a posteriori distribution for a channel source text given and a channel target text and a prior distribution for the channel source text =-=[1]-=-. Although there exists efficient algorithms to train the parameters for the statistical machine translation (SMT), one of the problems of the statistical machine translation is the search algorithms ... |

471 | Improved Statistical Alignment Models
- Och, Ney
- 2000
(Show Context)
Citation Context ...f this decoding algorithm. One of the strategies under consideration is a method to restrict the types of phrases, not only by language model probabilities, but also by other means, such as templates =-=[12]-=-, FSA-based approach [5] or [11]. The language model deployed for this experiment was a conventional n-gram, which estimates the likelihood of phrases by exploiting only left side context. Although th... |

248 | A maximum entropy approach to adaptive statistical language modeling. Computer speech and language - Rosenfeld - 1996 |

119 | Decoding complexity in word-replacement translation models
- Knight
- 1999
(Show Context)
Citation Context ...extremely large,|a| l+1 , where the l is the output length, an approximation of P(f|e)�P(f, a|e) is used when exploring the possible candidates of translation. This problem is known to be NP-Complete =-=[8]-=-, for the reordering property in the model further complicates the search. One of the solution is the left-to-right generation of output by consuming input words in any-order. Under this constraint, m... |

117 | Fast Decoding and Optimal Decoding for Machine Translation. ACL-01
- Germann, Jahr, et al.
- 2001
(Show Context)
Citation Context ...gies assuming linearly aligned translation source and target texts. The algorithms proposed above cannot deal with drastically different word correspondence, such as Japanese and English translation. =-=[7]-=- suggested greedy method and integer programming decoding, though the first method suffer from the similar problem as described above and the latter is impractical for the real-word application. The a... |

56 |
Language translation apparatus and method of using context-based translation models
- Berger, Brown, et al.
- 1996
(Show Context)
Citation Context ...ical machine translation (SMT), one of the problems of the statistical machine translation is the search algorithms of the translation given a sequence of words. There exists stack decoding algorithm =-=[2]-=-, A* search algorithm [3, 4] and dynamicprogramming algorithms [5, 6]. They all translate the given input string word-by-word and render the translation left-to-right, with pruning technologies assumi... |

40 |
Statistical machine translation, final report, JHU workshop. http://www.clsp.jhu.edu/ws99/projects/ mt/final report/mt-final-report.ps
- Al-Onaizan, Curin, et al.
- 1999
(Show Context)
Citation Context ...ndidates E f for each input words. The j translation candidates can be computed from the inverse translation probability derived by the Lexical Model and the distribution of e, as suggested by [2] or =-=[9]-=-. Given the inverse translation matrix, the candidates are derived from the top n most probable set of words. The algorithm generates candidate cepts with the fertility up to max f ertility from the l... |

39 | Decoding algorithm in statistical machine translation
- Wang, Waibel
- 1997
(Show Context)
Citation Context ...(SMT), one of the problems of the statistical machine translation is the search algorithms of the translation given a sequence of words. There exists stack decoding algorithm [2], A* search algorithm =-=[3, 4]-=- and dynamicprogramming algorithms [5, 6]. They all translate the given input string word-by-word and render the translation left-to-right, with pruning technologies assuming linearly aligned translat... |

27 | An efficient A* search algorithm for statistical machine translation
- Och, Ueffing, et al.
- 2001
(Show Context)
Citation Context ...(SMT), one of the problems of the statistical machine translation is the search algorithms of the translation given a sequence of words. There exists stack decoding algorithm [2], A* search algorithm =-=[3, 4]-=- and dynamicprogramming algorithms [5, 6]. They all translate the given input string word-by-word and render the translation left-to-right, with pruning technologies assuming linearly aligned translat... |

18 | Word re-ordering and dp-based search in statistical machine translation
- Tillmann, Ney
- 2000
(Show Context)
Citation Context ...ical machine translation is the search algorithms of the translation given a sequence of words. There exists stack decoding algorithm [2], A* search algorithm [3, 4] and dynamicprogramming algorithms =-=[5, 6]-=-. They all translate the given input string word-by-word and render the translation left-to-right, with pruning technologies assuming linearly aligned translation source and target texts. The algorith... |

17 |
Eiichiro Sumita, Fumiaki Sugaya, Hirofumi Yamamoto, and Seiichi Yamamoto. 2002. Toward a broad-coverage bilingual corpus for speech translation of travel converstaions in the real world
- Takezawa
(Show Context)
Citation Context ...m perplexity 26.16 36.92 4. EXPERIMENTAL RESULTS The corpus for this experiment consists of 172,481 bilingual sentences of English and Japanese extracted from a large-scale travel conversation corpus =-=[10]-=-. The statistics of the corpus are shown in Table 1. The database was split into three parts: a training set of 152,183 sentence pairs, a validation set of 10,148, and a test set of 10,150. The transl... |

4 | Search algorithms for statistical machine translation based on dynamic programming and pruning techniques
- Garcia-Varea, Casacuberta
(Show Context)
Citation Context ...ical machine translation is the search algorithms of the translation given a sequence of words. There exists stack decoding algorithm [2], A* search algorithm [3, 4] and dynamicprogramming algorithms =-=[5, 6]-=-. They all translate the given input string word-by-word and render the translation left-to-right, with pruning technologies assuming linearly aligned translation source and target texts. The algorith... |

4 |
Eiichiro Sumita, “Statistical machine translation based on hierarchical phrase alignment
- Watanabe, Imamura
- 2002
(Show Context)
Citation Context ...thm, presented above and beam search algorithm presented in [3] with the same thresholding parameters. The result was evaluated by word-error-rate (WER) and position independent word-error-rate (PER) =-=[11, 3]-=-. The WER is the measure by penalizing insertion/deletion/replacement by 1, and PER is the one similar to WER but ignores the positions, allowing the reordered outputs. Since the trigram perplexity is... |