W14-3316 a translation rule , including multi-word alignments . Again , class-based translation
C04-1005 single word alignment than on multi-word alignment . In Table 4 , the precision
W09-0411 substitution taking into account multi-word alignments instead of just single word mappings
C04-1005 Otherwise , we use the multi-word to multi-word alignment algorithm in Figure 3 to modify
C04-1005 our method is able to handle the multi-word alignments with higher accuracy , which
W07-0701 target nodes . In the case of multi-word alignments , all contiguous 3 aligned nodes
C04-1005 achieves much better results on multi-word alignment than other methods . However
C04-1005 First , we will further improve multi-word alignment results by using other technologies
P10-1085 a target word is a many-to-one multi-word alignment . To improve phrase table , we
P10-1085 error rate reduction of 29 % on multi-word alignment . The improved word alignment
C04-1005 " In - ter " because it has no multi-word alignment links . All of the methods perform
C04-1005 there is a word to word or word to multi-word alignment link i For Case 1 , we first
P10-1085 this paper significantly improves multi-word alignment , achieving an absolute error
C04-1005 source language and improve the multi-word alignment results . Experimental results
P10-1085 compared to single-word alignment , multi-word alignment is more difficult to be identified
P08-1113 Competitive Linking to deal with multi-word alignments and takes advantage of word-internal
P10-1085 that of the baseline method . For multi-word alignments , our methods significantly outperform
C04-1005 word to word alignment , word to multi-word alignment , multi-word to word alignment
P08-1113 linked at best . 4.1 Dealing with multi-word alignment We made a small change to Competitive
P10-1085 i.e. CM-3 , the error rate of multi-word alignment results is further reduced .
hide detail