Hello all, I am trying to score a set of rules that, even thought they do not come from any tree in either the source or the target language, they look like those used for string-to-tree SMT. An excerpt of the extract.sorted.gz file follows:
King Solomon [X][X0] [X] ||| [X][X0] rey salomón [X0] ||| 0-1 1-2 2-0 ||| 0.2 ||| King Solomon [X][X0] [X] ||| [X][X0] rey salomón [X0] ||| 0-1 1-2 2-0 ||| 0.333333 ||| King [X] ||| rey [X0] ||| 0-0 ||| 1 ||| Solomon [X] ||| salomón [X0] ||| 0-0 ||| 1 ||| [X][X0] mines [X] ||| las minas [X][X0] [X0] ||| 1-0 1-1 0-2 ||| 0.2 ||| [X][X0] mines [X] ||| las minas [X][X0] [X0] ||| 1-0 1-1 0-2 ||| 0.5 ||| When I try to score these rules to obtain the rule table I get the following error: ERROR: target phrase does not match in line 1: '' s [X]' != '' s [X0]' which is produced in the process of consolidating direct and indirect rule tables. Is there any way to overcome this problem other than editing the code to avoid the corresponding checking? PS: I saw a similar question in the archive, but with no answer; see http://moses-support.mit.narkive.com/r4NooMQR/score-externally-acquired-phrases Thanks you very much in advance!! Best wishes -- Felipe _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
