hi xiang
even if 'red' is not in your corpus, the source span [2-4] will still be
an NP. If you then had a rule
VP -> [V,1] [NP,2] ||| X -> [X,1] [X,2]
it can still apply to span [1-4].
On 18/08/2010 12:51, ?????? wrote:
Hi,
for example,when I use a Tree-To-String model to decode the English
sentence "I eat apple."
and the input format is:
<tree label="NP" span="0-0"/> <tree label="VP" span="1-4"/> <tree
label="V" span="1-1"/> <tree label="NP" span="2-4"/> I ate a red apple .
if the word "red" has not yet occur in the training data,so it can't
be denoted as "ADJ",in this case, the string "a red apple" will still
be denoted as "NP" since the input indicates it ,or be denoted as "X"
since the manual points out that "X" is prepared for unknown word.
Best regards
Xiang
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
