Methodology for the Construction of Multi-Structured Documents

Multi-Structured Documents (1)

Multiple uses of a same document

Multiplication of documentary structures

• Physical
• Logical
• Semantic ...

SGML, XML: overlapping hierarchies

Last Minute (1)

"The Structure Of Appearance"

The Appearance Of Structure

Last Minute (2)

Aporia:

• Every Document is a String
• A String cannot be Modified
• A Document can be Modified

A document has no structure

=> The "String-In-A-Role" strategy

Multi-Structured Documents (2)

Working environment

Institut Jean-Toussaint Desanti

• Philosopher

• Main work: "Les Idéalités Mathématiques"

  Development of the theory of real variables functions

Multi-Structured Documents (3)

ZI: meaningful fragment of textual content spanning two pages

S1: pages

S2: "regions of interest"

E1, E2: equations

2 or 3 structures ...

Representation of M-S.D. (1)

4 categories

• Historical solution: CONCUR
• Ad-Hoc solutions
• Non XML-compatible models
• XML-compatible models

Representation of M-S.D. (2)

5 dimensions

• Expressiveness
• Genericity
• Quality of implementation
• Query mechanisms
• Change management

Representation of M-S.D. (3)

CONCUR

• Multiple DTDs for the same content
• Prefixed tags
• No relations between structures
• Tags order may be indeterminable
• No query mechanism

MuLaX: adaptation of SGML CONCUR to XML

Representation of M-S.D. (4)

Ad-Hoc solutions

• Duplication of content
• Empty tags
• A game of well chosen attributes
• Isolation of the content in a base structure

Representation of M-S.D. (5)

TexMECS and LMNL

• definition of new syntaxes
• LMNL: interval algerba ...

Representation of M-S.D. (6)

Annotation graphs

Have been developed to model linguistics phenomena.

RDF graphs: annotation graphs in a well known formalism

Last Minute (3)

EARMARK: Extreme Annotations RDF Markup

Representation of M-S.D. (7)

GODDAG

• Graph model
• XPath extension
• Relations between elements of distinct structures
• Import/Export from/to XML

Representation of M-S.D. (8)

MCT (Multi-Colored Trees)

• Extension of the XML model
• Tree coloring
• A node may have multiple colors
• Extension of XPath step to navigate between colors
• An XUpdate extension should be possible

Representation of M-S.D. (9)

MSXD (Multi-Structured XML Documents)

• Annotation of structures
• Introduction of a MSD schema
• XQuery extension

Representation of M-S.D. (10)

Delay Nodes

• Extension of the XDM model
• Delay node == virtual representation as an XQuery query of some of the child node of their parent node
• Model very similar to the one of MCT (documents as sets of XML trees)

Representation of M-S.D. (11)

MonetDB/XQuery

• XML SGBD
• Stand-Off markup technique
• Optimized query operators

Representation of M-S.D. (11)

MSDM: a document is a graph D composed of:

• A set of nodes BS (Base Structure)
• A family (DS_j) with j ∈ J of trees (Documentary Structures)
• ∀ j ∈ J ∃ a relationship R_j that associates each node of (DS_j) with a subset of BS
• For each leaf of (DS_j) this subset is non empty.

Representation of M-S.D. (12)

MultiX, XML instance of MSDM (1)

First Structure:

<s1>
 <page>Autrement dit la distinction signe-signifie ...
   Remarque,
 </page>
 <page>ce discours, ...
   par ex le discours 3+2=0-1 est-il un texte ? ...
 </page>
</s1>

MultiX, XML instance of MSDM (2)

Second Structure:

<s2>
<p>Autrement dit la distinction signe-signifie...</p>
<p>Remarque, ce discours, ...</p>
<p>par ex le discours
 <eq>3 + 2 = 0 - 1</eq> est-il un texte ? ...</p>
</s2>

MultiX, XML instance of MSDM (3)

Base Structure:

<seg xml:id="F1">Autrement dit la distinction
signe-signifie ...</seg>
<seg xml:id="F2">Remarque, </seg>
<seg xml:id="F3">ce discours, ...</seg>
<seg xml:id="F4">par ex le discours </seg>
<seg xml:id="F5">3 + 2 = 0 - 1</seg>
<seg xml:id="F6"> est-il un texte ? ...</seg>

MultiX, XML instance of MSDM (4)

First Structure:

<s1>
<page>
 <xi:include href="b.xml" xpointer="element(F1/1)"/>
 <xi:include href="b.xml" xpointer="element(F2/1)"/>
</page>
<page>
 <xi:include href="b.xml" xpointer="element(F3/1)"/>
 <xi:include href="b.xml" xpointer="element(F4/1)"/>
 <xi:include href="b.xml" xpointer="element(F5/1)"/>
 <xi:include href="b.xml" xpointer="element(F6/1)"/>
</page>
</s1>

MultiX, XML instance of MSDM (5)

Second Structure:

<s2>
<p>
 <xi:include href="b.xml" xpointer="element(F1/1)"/>
</p>
<p>
 <xi:include href="b.xml" xpointer="element(F2/1)"/>
 <xi:include href="b.xml" xpointer="element(F3/1)"/>
</p>
<p>
 <xi:include href="b.xml" xpointer="element(F4/1)"/>
 <eq>
  <xi:include href="b.xml" xpointer="element(F5/1)"/>
 </eq>
 <xi:include href="b.xml" xpointer="element(F6/1)"/>
</p>
</s2>

MultiX, XML instance of MSDM (6)

XQuery functions:

let $physique := doc("physique.xml")
let $logique := doc("logique.xml")
for $page in $physique//page,
    $para in $logique//p
where multix:share-fragments($page,$para) and
      not(multix:include-content-of($page,$para))
return $para

Finds region of interest overlapping two pages.

Construction of M-S.D. (1)

Three categories of methods:

• Automatic differentiation of structures
• Confrontation of a user with the results of automatic restructuring
• Creation of a network of documents authors

Construction of M-S.D. (2)

Restructuting stage

Pages and regions of interests have been tagged until a region overlaps two pages.

Construction of M-S.D. (3)

Automatic restructuting

Creation of a new structure is a purely formal operation consisting in the transformation of a graph into two trees.

Construction of M-S.D. (4)

Integration of the user

Automatic restructuring is a good occasion for a user to make modeling choices.

Construction of M-S.D. (5)

Recommendation system for documents authors

Two users are close, insofar as they are editing specific documents, if the implied tags trees of their structures are close.

Prototype implementation (1)

REST interface

ex: POST to http://desanti.org/collections/148/structures/math/taggees with content:

<taggee>
  <tag name="equation" />
  <interval start="14" end="26" />
</taggee>
</programlisting>

Prototype implementation (2)

Javascript user interface

Conclusion

Multi-Structured documents: from their representation to their construction

The enforcement of tree structures, for a long time considered as the crux of the M-S.D. problem, triggers the creation of new structures that have to be validated by the user.

We propose an open methodology that can be used for the incremental and collective emergence of documentary structures.

Questions