exclusion contexts in patterns

[unhammer]

I want lone determiners/adjectives to be antecedents. In Norwegian, we get these ellipses like the<mf><sg> was sold or red<pl> was drawn – both of those are good antecedents, but we never want the determiner in the car or adjective in red paint to be an antecedent. So it'd be nice to be able to say

      <pattern>
        <pattern-item n="det"/>
        <negate><pattern-item n="noun_or_adj"/></negate>
      </pattern>

and have that match determiners not followed by noun or adj.

---

Alternatively, would it be easier to add a score per pattern-item? Then I could make a negative rule for modifying determiner/adj's of noun phrases, e.g.

    <markable n="DETMOD"> <!-- score modifying determinatives lower (we only want *lone* determinatives to be antecedents) -->
      <pattern>
        <pattern-item n="det" score="-1"/>
        <pattern-item n="adj" score="-1"/>
        <pattern-item n="nom" />
      </pattern>
</markable>

or something like that.

[khannatanmai]

Hey,
Both of these are good ideas and prolly should be implemented. The part about individual scores to elements was something I was going to implement, just never found a use case for it.

As for the first part, I coded this thing more than two years ago so I'm not very sure, but I was just looking in the documentation:
def-parameters has this element called "exclude-tags" in addition to "has-tags". Now there's no example of this in the documentation but if we have the "exclude-tags" element in def-cats as well, you could do something like:

<def-cat n="det">
	<cat-item has-tags="det"/>
	<cat-item has-tags="det pos"/>
</def-cat>

<def-cat n="not_noun_or_adj">
        <cat-item exclude-tags="n"/>
	<cat-item exclude-tags="adj"/>
</def-cat>

and then in markable:

<pattern>
        <pattern-item n="det"/>
        <pattern-item n="not_noun_or_adj"/>
</pattern>

I do recognise that this would be more tedious than just being able to negate pattern-items, plus I haven't even tried this yet. So can look into it.

[unhammer]

Ah, yeah that might help :-) didn't know you could have just exclude without has-tags :)

[khannatanmai]

I'm not sure either tbh :p try it once

…

On Mon, Jun 27, 2022, 01:45 Kevin Brubeck Unhammer ***@***.***> wrote: Ah, yeah that might help :-) didn't know you could have *just* exclude without has-tags :) — Reply to this email directly, view it on GitHub <#50 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGJBIHQYBTEFM5DA2LYH5FDVRC26TANCNFSM5ZY6YEKQ> . You are receiving this because you commented.Message ID: ***@***.***>

[unhammer]

That does work! =D