put scope on ref to indicate where the reference applies

[ttasovac]

this would complementary to ref type... i#m just parking this issue here, laurent and i will elaborate

[laurentromary]

e.g. <ref type='lemma' scope='currentEntry'/>

[ttasovac]

The following is a proposal that's based on Ana's, Rute's and my presentation at the collocation workshop in Sintra + subsequent discussions with Laurent.

We got rid of oRef and we don't want to resurrect it. We have a robust mechanism for cross-references (xr/ref) where xr is the container, and the actual reference string is in ref + we can add things like lbl etc. A while back we added types to ref (entry, sense) to distinguish the type of reference (whether it's pointing to an entry as a whole, or a particular sense). So far, so good.

At the same time, we need a mechanism for collocations (i.e. "transparent", undefined MWEs...), like here:

In + da fonte, da torneira, plus is a ref to the lemma, it is used only in the first collocation, and in the second it is only implied.

<sense xml:id="agua.4">
    <!--etc.-->
    <form type="collocations">
        <form type="collocation">
            <orth>
                <xr type="include">
                    <lbl>+</lbl>
                    <ref type="lemma" scope="currentEntry" target="#agua.lemmaid"/>
                    <seg>da fonte</seg>
            </orth>
        </form>
        <pc>,</pc>
        <form type="collocation">
            <orth>
                <xr type="include">
                    <ref type="lemma" scope="currentEntry" target="#agua.lemmaid"/>
                    <seg>da torneira</seg>
                </xr>
            </orth>
        </form>
    </form>
    <pc>;</pc>
    <!--etc.-->
</sense>

The thinking behind it is:

• colloctions of this kind definitely need to be encoded as form/orth
• lexicographic shorthand (+, ~, implied refs) need to be resolvable automatically, i.e. we need to encode things in such a way that we can extract full collocations if we wanted
• we already have a mechanism for cross-references, so we should really use it here as well
• we should discuss the possibility of introducing @scope to <ref> and then agree on a typology (a la: currentEntry, currentSense, externalEntry, externalSense etc...) to highlight where we are pointing to; this clearly needs more discussion
• seg is the default inside orth for non-ref strings. of course, one could turn everything into refs, if one wanted to encode every word as a link

Some issues:

1. see #64

2. I would like to have the option of having @valueon <ref> so that one can explicitly put in the value of the "missing" lemma (i.e. "resolve" the plus sign inline) should one want. The definition of @value is quite appropriate for this, but it's not allowed on ref.

3. We need to figure out and agree on how to type forms here (form type = "collocations", form type="collocation", which we will also have to consider together with #65.

[ttasovac]

@laurentromary

we suggest to use @expand instead of @value, once we add ref to att.lexicgoraphic