Exploring%20intertextual%20semantics

1
Exploring intertextual semantics

• A reflection on attributesand optionality
• Yves MARCOUX Élias RIZKALLAH
• GRDS EBSI
• Université de Montréal

2
Structure of the talk

1. NL-based modeling and intertextual semantics (IS)
   (quick review)
2. IS hypotheses (ISH-1 and ISH-2)
3. Attributes in the light of IS
4. Optional containers in the light of IS
5. Conclusion
6. Question period

3
Intertextual semantics (1/2)

• Usual semantics
• Intertextual semantics

Natural orartif. lang.
Formal domainex. 1st order logic
S
Artificiallanguage
S
Natural language
4
Intertextual semantics (2/2)

• A framework in which we can give a precise
  meaning in NL to the artificial constructs that
  are structured documents
• Is the underlined part an oxymoron?
• Not from some standpoints
• Wittgenstein (I think!)
• Precision up to a level deemed adequate by the
  document architect (modeler)
• For some given target community

5
NL-based modeling (1/2)

• Modeler works out structural declarations and
  corresponding IS simultaneously
• EML2006 explored minimalist approach two
  peritexts for each element type
• One text-before segment
• One text-after segment
• Both are fixed (constant)
• Hyperlinks are allowed in peritexts

6
NL-based modeling (2/2)

• Modeler starts with desired IS (prose)
• Identifies structural regularities and
  corresponding peritexts in the prose
• Works out structural declarations
• Chooses element names as abbreviations of
  peritexts

7
A small example

• Raw XML

ltbillinggt ltamount-burialgt1205.47lt/amount-burialgt
ltpayable-burialgtDlt/payable-burialgt
ltamount-cremationgt788.00lt/amount-cremationgt
ltpayable-cremationgtFlt/payable-cremationgtlt/billing
gt
8
IS specification
Element text-before text-after
billing "This section gives the billing information for this order. " " End of billing information section."
amount-burial "Amount charged for the burial service " " canadian dollars  "
payable-burial "this amount is payable by " " (D Funeral director F Family)."
amount-cremation "Amount charged for the cremation service " " canadian dollars  "
payable-cremation "this amount is payable by " " (D Funeral director F Family)."
9
IS view
10
What use can this have?

• Mostly an upstream semantics
• Give easily understandable meaning to a document
  being created
• Facilitates modeler-author communication
• But possible downstream uses
• reference interpretation (legal?) when multiple
  renderings are possible
• text-only version for text-only tools (ex.
  full-text indexing)

11
More fundamentally

• An IS specification captures the essence of what
  structural constructs are made for
• XML elements / attributes
• RDB tables (The ltfield-namegt is ... .)
• Helps in picking up good names
• Clearly reveals the ease / difficulty of making
  sense of our documents

12
Has links with...

• Literate programming
• Computer documentation

13
Possible avenuesas of EML2006

• Implementation
• IS view / expansion in an XML editor
• Apply to existing models
• Popular ones RSS, Atom, DocBook, TEI Lite
• Explore how to treat attributes
• Peritexts associated to element types only
• Make explicit the hypotheses under which all this
  is interesting...

14
Structure of the talk

1. NL-based modeling and intertextual semantics (IS)
   (quick review)
2. IS hypotheses (ISH-1 and ISH-2)
3. Attributes in the light of IS
4. Optional containers in the light of IS
5. Conclusion
6. Question period

15
ISH-1 (1/2)

• The IS of a document corresponds to its meaning
  (intended interpretation)
• Assumed to be understandable (interpretable) by
  the (human) members of the target community
  either directly, or
• by navigating through the network of resources
  anchored (via hyperlinks) in the IS of the
  document

16
ISH-1 (2/2)

• The network needs only extend until it reaches
  resources directly understandable by the members
  of the target community
• This network of resources suggests an actual
  interpretation (sense-making) path, but does not
  impose it
• Any specific reading of a document yields more
  information than the IS, but IS is a guaranteed
  minimum for all readings

17
ISH-1 -- Example

• Documentltamount currency"MNT"gt100lt/amountgt
• ISAmount (in the currency whose ISO 4217 code
  is "MNT") 100 (the table of ISO 4217 codes can
  be consulted at http//www.iso.org/iso/en/prods-s
  ervices/popstds/currencycodeslist. html)

18
ISH-1 -- Notes

• Intricate structure of the IS prose,many
  hyperlink traversals needed gtdocuments are hard
  to understand
• Semantic precision
• Evolution, vagueness, are not a problem
• Mathematical precision neither

19
ISH-2

• A raw XML document should look like an
  abbreviation of its IS
• Thus, element names (generic IDs) should look
  like abbreviations of their respective peritexts
• Related to information perennity

20
Structure of the talk

1. NL-based modeling and intertextual semantics (IS)
   (quick review)
2. IS hypotheses (ISH-1 and ISH-2)
3. Attributes in the light of IS
4. Optional containers in the light of IS
5. Conclusion
6. Question period

21
What are attributes, really?

• Subelements?lta href"dog.jpg"gtClicklt/agtis
  simply an abbreviation forltagtlt_at_hrefgtdog.jpglt/_at_h
  refgtClicklt/agt

22
Not quite...

• Intrinsic lack of ordering
• (Name conflicts)
• Semantic interplay between element and attribute
  (tweaking)
• Can we account for this in IS?

23
Place-holders in peritexts

• Text-before for element amount"Amount (in the
  currency whose ISO 4217 code is _at_currency) "
• IS of example becomesAmount (in the currency
  whose ISO 4217 code is MNT) 100.

24
Structure of the talk

1. NL-based modeling and intertextual semantics (IS)
   (quick review)
2. IS hypotheses (ISH-1 and ISH-2)
3. Attributes in the light of IS
4. Optional containers in the light of IS
5. Conclusion
6. Question period

25
What if the attribute is omitted?

• Treat as empty value?
• IS of example becomesAmount (in the currency
  whose ISO 4217 code is ) 100.
• Ill-formed gt bad!

26
Marked sections in peritexts

• Text-before for element amount"Amount_at_currency
  (in the currency whose ISO 4217 code is _at_) "
• IS of example (with attribute omitted) Amount
  100.

27
Its OK to omit, but...

• It should be clear (somewhere in the IS) that
  there could have been something at the point of
  omission
• What it is that could have been there
• Why? Because all that contributes to making sense
  of the document

28
Structure of the talk

1. NL-based modeling and intertextual semantics (IS)
   (quick review)
2. IS hypotheses (ISH-1 and ISH-2)
3. Attributes in the light of IS
4. Optional containers in the light of IS
5. Conclusion
6. Question period

29
All this is meant to...

• Help modelers (and developers) not forget that,
  want it or not, users do assign meaning to
  documents
• If at least one sense making path is not devised
  by the modeler, users will roll their own, with
  possibly unpredictable results
• Allow developers to specify and useably deliver
  to users such sense making paths

30
Future work includes

• Apply to existing models
• More powerful peritext computation?
• Need more than peritexts hyperlinks?
• Need more than NL? Images? Sounds?
• Multilingualism
• Apply to other models / interface design
• Experiment in authoring situations

31
Thank you!

• Questions?