[developers] Valid MRS? Bug in ERG?

[goodman.m.w at gmail.com]

On Fri, Sep 25, 2020 at 12:02 PM Alexandre Rademaker <arademaker at gmail.com>
wrote:

> Hi Michael,
>
> Thank you for your comments.
>
> The main fact now is that LKB does not produce  disconnected MRSs for the
> same sentence.  Maybe ACE error or a difference on the initialization
> scripts of ERG for each tool? I understood the possible difference in the
> approach for converting MRS to EDS between LKB and PyDelphin, but the MRSs
> are different to begin with.
>

Oh I see. I missed that part of your message. Different parse-ranking
models, perhaps?


> Good point about the incomplete DMRS output.   One more hasty conclusion
> from my side. I didn’t inspect carefully the DMRS produced.
>
> How can I tell edm to use DMRS instead of EDS? Maybe I missed
> something...
>

The -f / --format option specifies the codec to use. E.g., --format=dmrx

Maybe edm must be more robust to ignore pairs with errors?
>

That's a good idea. I created
https://github.com/delph-in/delphin.edm/issues/3



> Alexandre
> Sent from my iPhone
>
> On 25 Sep 2020, at 00:32, goodman.m.w at gmail.com wrote:
>
> And regarding https://github.com/delph-in/delphin.edm, note that this
> implementation also works for MRS, by converting to EDS along the way, and
> for DMRS, without conversion.
>
>

-- 
-Michael Wayne Goodman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20200925/1b43dab6/attachment.html>