[developers] ERG paraphrasing with ACE

[Dan Flickinger]

Hi Francis -

As an example of the use of ACE and the ERG in such a paraphrasing setup, you might look at the `openproof' subdirectory, where I defined a set of paraphrase rules for the logic-education application, and use the following pipeline:

echo "B is a cube and C is a cube." | ace -g erg.dat -f | ace -g openproof/paraphrase-op.dat --transfer-config openproof/rules.all | ace -g erg.dat -e > /tmp/sents.txt

 Dan


----- Original Message -----
From: "Francis Bond" <bond at ieee.org>
To: "Woodley Packard" <sweaglesw at sweaglesw.org>
Cc: developers at delph-in.net
Sent: Thursday, January 29, 2015 5:23:58 PM
Subject: Re: [developers] ERG paraphrasing with ACE

G'day,


> ACE has a couple of mechanisms for automatically applying transfer rules, but they only work with unambiguous transfer rule sets.  You can load transfer rules with tdl status "fixup-rule" to make them apply before generation or tdl status "cleanup-rule" to make them apply after parsing (and also after generation, though most people don't look at the MRS that is generated with a realization).  I don't think this will work for  you though because when the rules are ambiguous, the system just takes the first result (and prints a warning).

Thanks for the explanation.

> So, assuming you want fan out, you need to put a separate transfer grammar phase in the middle.

Then we will do that.

> Regards,
> Woodley
>
> On Jan 29, 2015, at 7:16 AM, Francis Bond <bond at ieee.org> wrote:
>
>> G'day,
>>
>> Does anyone know if there is any simple way to
>> parse+paraphrase+generate using the ERG and ACE using the
>> paraphraser.mtr?
>>
>> i.e. to pass all out put through the paraphraser by default?  If not,
>> we can just build a light enen system, but I thought I'd see if there
>> was a short cut :-).
>>
>> --
>> Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
>> Division of Linguistics and Multilingual Studies
>> Nanyang Technological University
>



-- 
Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
Division of Linguistics and Multilingual Studies
Nanyang Technological University