[CWB] copying alignments

Ruprecht von Waldenfels ruprecht.waldenfels at gmx.net
Wed Dec 16 17:50:44 CET 2015


Stefan,
thanks! It turned out to be a stupid mistake (I had forgotten to update 
the data path in the registry file). But your answer gave me the 
confidence that the approach should work, and so the patience to look 
for the obvious solution...

Just to clarify: So that means once the alignment is computed and stored 
in the alx files, it becomes independent from the structural attribute 
that was used to define it?

Thanks,
Best,


  16.12.2015 um 00:31 schrieb Stefan Evert:
>> I have the same corpus two times, with different annotations, including different alignments.
> To be clear, these are two versions of the same corpus with exactly the same tokenization, but different annotations?  Then it should be ok simply to copy the alignment files.
correct
>
>> Now I wanted to copy an alignment from one to the other, i.e., add it to the other corpus. What I did is I simply copied the rng and alx files, and added a line to the registry (ALIGNED corpus).
> The alignment attribute only consists of the .alx file, so you don't need to copy any .rng ones unless you're also using them for some other purpose.
>
>> However, that did not work. Would it be expected to? How could one do such a thing? It would save me quite a bit of time.
> What does "cwb-describe-corpus -s" show on the second corpus. And what exactly do you mean by "did not work"?
>
> As a wild guess, I'd suppose that the second corpus is aligned to a different version of the target corpus under a different CWB name.  Then you have to adjust the name of the alignment attribute and the corresponding index file accordingly, of course.
>
> Generally, a safer strategy is to cwb-align-decode the alignment attribute, change the corpus names in the header line, and cwb-align-encode it for the new corpus.
>
> Best,
> Stefan
>
>
>
> _______________________________________________
> CWB mailing list
> CWB at sslmit.unibo.it
> http://devel.sslmit.unibo.it/mailman/listinfo/cwb



More information about the CWB mailing list