[developers] Structured output from ACE

[Woodley Packard]

Hi again Mike,

I’ve checked in a new option which produces UDX-formatted @type tags.  Using "--udx" produces tags just on the lexeme nodes, and "--udx=all" produces tags on the rule nodes too.

The UDX headedness marker ("^") is not generated at the moment.  The contents of the file that specifies headedness of rules (etc/rules.hds) is not stored in the compiled grammar image currently, but more critically, that format change breaks ACE’s assumption that partial derivation tree strings can be reused in different contexts (they are cached for faster generation of UDF when exhaustive or top-N unpacking is invoked).  It would be possible to work around that assumption, but I’m not going to do it without a clear need. :-)

--Woodley

> On Jun 24, 2016, at 11:59 AM, Michael Wayne Goodman <goodmami at uw.edu> wrote:
> 
> I will put a little more thought into whether and how to get the type
> information out to you.
> 
> Following up on my and Stephan's previous messages, the ItsdbDerivations wiki only mentions adding types on preterminal nodes, but its method of doing so (joining entity names and type names with @) doesn't change the tree structurally, so I imagine that method could easily be reused to allow the types on other nonterminals as well.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20160628/0eb1b593/attachment.html>