hi Shuoyang

thanks. I don't know enough about that script to commit your changes, but if you're sure about it, I can give you commit access for you to do it yourself, as long as you answer bug queries from other users. Or if anyone else knows more about it, maybe they can commit it for you.

analysis.perl just parses the debugging output, which will changes from time-to-time. I would worry too much if it says it's old or new format. Personally, i would only write it for the current debugging output format.

On 30/03/2015 02:59, Shuoyang Ding wrote:
Hi all,

I just upgraded from an older release (2.1.1) to the current version of moses. And I found the regular expression in line 780-781 and in line 1198-1204 of /scripts/ems/support/analysis.perl does not match the .trace and .graph files, respectively.

The .trace file I get is something like this:

Trans Opt 0 [0..12]: [12..12]=</s> [0..11]=S : S ->S -> S </s> :0-0 : term=1-1 : nonterm=0-0 : c=-0.219483 core=(0,-1,1,0,0,0,0,0,0,0) -210.72core=(-100,-22,14,-20.8622,-22.1086,-15.6026,-26.7554,-2.71614,0.999896,-101.045)

the .graph file:

0 40079 S => S -> <s> :: term=0-0 : nonterm=: c=0.110587 core=(0,-1,1,0,0,0,0,0,0,0) [0..0] [total=0.110587] core=(0,-1,1,0,0,0,0,0,0,0)

namely, the original regular expression seems fail to account for the “term” and “nonterm” stuff in both cases.

I attempted to fix the problem and I attached the fixed analysis.perl. I no longer get the error output again but I don’t know if it is the right way to do this, so I think it would be nice if some of you can look at that.

Plus, I’m a bit puzzled by the “old format” and “new format” indicated in function hs_scan_line. It seems that my experiment is matching to the “old format”, so is this old one deprecated? How can I switch to the new one?

Thanks!

Regards,
Shuoyang Ding

Ph.D. Student
Center for Language and Speech Processing
Department of Computer Science
Johns Hopkins University

Hackerman Hall 225A
3400 N. Charles St.
Baltimore, MD 21218

http://cs.jhu.edu/~sding <http://cs.jhu.edu/%7Esding>





_______________________________________________
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

Reply via email to