| C96-1037 | algorithm can produce effective | word-alignment | results with 1 . Read a pair |
| D08-1033 | increase in BLEU score over the | word-alignment | based heuristic estimates used |
| D08-1066 | probabilities and starts out from the | word-alignments | . The novel aspects of our model |
| A94-1006 | on part-of-speech tagging and | word-alignment | technologies to extract candidate |
| D13-1205 | type MISC ( miscellaneous ) . The | word-alignments | suggest we should transfer this |
| A94-1006 | for part-of-speech tagging and | word-alignment | algorithms . Although the output |
| D09-1075 | . We also used the most simple | word-alignment | model , but more complex word |
| D09-1108 | and Ney , 2003 ) to do m-to-n | word-alignment | and adopt heuristic " grow-diag-final-and |
| D08-1066 | al. , 2008 ) ) showing that most | word-alignments | of actual parallel corpora can |
| D10-1044 | heterogeneous nature of SMT components ( | word-alignment | model , language model , translation |
| D12-1021 | it was combined with a standard | word-alignment | initialisation , thus leaving |
| D08-1060 | triggered a huge uproar in the local . | word-alignments | , our rules learned are still |
| C04-1154 | alternatively , manually established | word-alignments | to intiate the automatic substructural |
| C02-1160 | bi-texts to automatically create | word-alignments | ; in many statistical MT systems |
| D11-1046 | system developers in obtaining good | word-alignment | performance off-the-shelf when |
| D08-1066 | does not need to generate the | word-alignments | explicitly , as these are embedded |
| D14-1176 | dependency parse . Phrase-internal | word-alignments | , needed to segment the translation |
| D13-1205 | supervised source - side model and the | word-alignment | filtering techniques ( steps |
| D09-1136 | approach are the very reasons that | word-alignments | are used for rule extraction |
| D08-1033 | relative to the highly developed | word-alignment | - centered baseline , we show |
