RE: [xliff-inline] Sample for metadata inline representation

[Yves Savourel <ysavourel@enlaso.com>]

Here is a first shot.
In this representation I'm going for one generic element: <mrk>

It would have the following attributes:

- a mandatory 'type' attribute with the following values: trans, notrans, term, comment. And allow custom values in the form "x-value" (or maybe "xyz:value" to go in the direction of using namespaces for values)

- an optional 'idRef' attribute that points to an element inside the <unit> that hold additional information about the span of content associated with the given type.

- (possibly) an optional 'id' attribute that would allow elements in the <unit> to point to a given <mrk>. The id value would be the same in source/target when the <mrk> exists in both.

The main drawback of this, I think, is that any extra information beyond the type has to be fetched from outside the content. Probably not a big issue as long as all in kept in the unit.


<unit id='1'>
 <segment>

  <source>It would certainly be quite a <mrk type='notrans'>faux pas</mrk> to start a dissertation on the origin of <mrk type='term' idRef='t1'>modern novel</mrk> without mentioning <mrk type='note' idRef='c1'>the <mrk type='xyz:book-title' idRef='cust1'>Epic of <mrk type='abc:proper-noun'>Gilgamesh</mrk></mrk>. </mrk></source>

  <target>Ce serait certainement un <mrk type='notrans'>faux pas</mrk> grave de commencer une thèse sur <mrk type='note' idRef='c2'>l'origine du <mrk type='term' idRef='t1'>roman moderne</mrk></mrk>, sans mentionner <mrk type='note' idRef='c1'>l'<mrk type='xyz:book-title' idRef='cust1'>Épopée de <mrk type='abc:proper-noun'>Gilgamesh</mrk></mrk>. </mrk></target>

</segment>
 <segment>

  <source><mrk type='note' idRef='c1'>There is also the <mrk type='xyz:book-title' idRef='cust2'>Story of <mrk type='abc:proper-noun'>Sinuhe</mrk></mrk>.</mrk></source>

  <target><mrk type='note' idRef='c1'>Il y a aussi l'<mrk type='xyz:book-title' idRef='cust2'>Histoire de <mrk type='abc:proper-noun'>Sinouhé</mrk></mrk>.</mrk></target>

 </segment>

 <termDefs>
  <termDef id='t1'>Info about the term 'modern novel'</termDef>
 </termDefs>

 <comments>
  <comment id='c1' origin='author'>Comments about the two cited stories.</comment>
  <comment id='c2' origin='translator'>Translator's comments.</comment>
 </comments>

 <xyz:annotation id='cust1' type='book-title'>Some info about the Epic of Gilgamesh</xyz:annotation>
 <xyz:annotation id='cust2' type='book-title'>Some info about the Story of Sinuhe</xyz:annotation>

</unit>


Marker by marker:

--a) "faux pas" should be marked as not-to-be-translated.

I've used <mrk type='notrans'>, we would have also a type='trans' to indicate a span should be translated.


--b) "modern novel" ("roman moderne") should be marked as a term and have some associated information/definition.

I've used <mrk type='term' idRef='t1'> so the element 't1' hold the term-specific metadata, like definitions, source, example, reference to more info, etc.


--c) "Epic of Gilgamesh" ("Épopée de Gilgamesh") should have a reference to a custom information indicating it's a "book title" and holding information about that work.

I've used <mrk type='xyz:book-title' idRef='cust1'>. The value of type is custom, using a namespace prefix to allow more flexibility than just a 'x-value' pattern. (at the same time, this allows to have different values in different namespace to mean the same things, which is not too good...)


--d) "Story of Sinuhe" ("Histoire de Sinouhé") should have a reference to a custom information indicating it's a "book title" and holding information about that work.

Same as c).


--e) "the Epic of Gilgamesh. There is also the Story of Sinuhe." ("l'Épopée de Gilgamesh. Il y a aussi l'Histoire de Sinouhé.") should be a span referring to an author comment about both stories.

I've used <mrk type='note' idRef='c1'> to point to a <comment> element in that <unit>. Note that we would use two mrk, one in each segment, and if the segment is merge I assume we would have the processing expectation that adjacent markers with the same information are merged.


--f) "Gilgamesh" should be marked up as a "proper noun".

I've used <mrk type='abc:proper-noun'> assuming there is a namespace 'abc' that provides part-of-speech values.


--g) "Sinuhe" ("Sinouhé") should be marked up as a "proper noun".

Same as f).


--h) "l'origine du roman modern" should be associated with a translator comment.

I've used <mrk type='note' idRef='c2'>, the origin would be define in the element we are pointing to. We would have the requirement for the container to have some origin information and provide a list of the values.
I'm not too sure about this. It's hard to see where the inline markup information should stop.


That's all for now.
-yves