Training fulltext - annotation doubt

[martasoricetti]

I’m training the full text model and I have a doubt. It could happen that a reference is divided by a figure, a table or a formula.
How should I handle this situations in the annotation process?

<p>[…]<ref type="biblio">(Ramanathan et<lb/></ref></p>
 
 <figure type="table">Table 5. […] </figure> 
 
 <figure>Figure 3. […] </figure>
 
 <p><ref type="biblio">al., 2001)</ref>. However, it was argued that the use of APCADA,<lb/>[…]</p>

I tried this approach (dividing the same intext reference pointer in two different tags) but i don’t know if it’s the right choice

[lfoppiano]

HI @martasoricetti, I think that's the right way. However I'm not sure how those rare cases will be reconstructed after the model extracts them. But yes, that's the right approach.

You could add xml:id / corresp attributes only to those references that are split by other elements.
Those attributes will be ignored at the moment, but they might be used in future to establish that it's the same reference.

Something like:

<p>[…]<ref type="biblio" xml:id="ref1">(Ramanathan et<lb/></ref></p>
 
 <figure type="table">Table 5. […] </figure> 
 
 <figure>Figure 3. […] </figure>
 
 <p><ref type="biblio" corresp="#ref1">al., 2001)</ref>. However, it was argued that the use of APCADA,<lb/>[…]</p>