Other recent changes to the scripts have enhanced tokenizer.perl to 
 handle non-printing ASCII and other control characters. A few weeks ago, 
 the section #escape special chars (~ line 151) updated to include 
 ' and ".

 At that time, the square brackets [] were escaped to &bra; and &ket;. 
 Now, I see they're escaped to [ and ]. This is great 
 because it eliminates two custom named entities "bra" and "ket". I'd 
 like to suggest the same for the vertical bar escaping, which is 
 currently escaped to &bar;, and change it to |. This change 
 would eliminate the remaining custom named entity. Detokenize.perl would 
 need updating to include the new escaping and add &bar; to the legacy 
 section.

 (resent because my browser mail client converted the escaping)
 Tom


 On Fri, 22 Jun 2012 09:26:17 +0100, Hieu Hoang 
 <[email protected]> wrote:
> ah, that's where the mgizapp name comes from. I'll ask qin to make 
> the
> automake & cmake exec names consistent (ie. mgiza).
>
> On 22/06/2012 05:53, Tom Hoar wrote:
>> Hieu,
>>
>> Another observation about the changes. Compiling the new MGIZA++ 
>> with
>> cmake on my system yesterday created the binary $prefix/bin/mgizapp.
>> Line 228 of the most recent train-model.perl reads: $GIZA =
>> "$_EXTERNAL_BINDIR/mgiza"
>>
>> The current/valid binary name of MGIZA++ has been a topic of
>> discussion here before. However, since the file name changes
>> unexpectedly between builds and this topic is "changes to the 
>> training
>> scripts", would it be better if train-model.perl tests for both
>> possible filenames for MGIZA++ binary?
>>
> _______________________________________________
> 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