TEI TO RDF

Premise:

• TEI-XML has many limitations in the representation of the "enriched text".
• The XML element encapsulates a node's properties as attributes, but (unlike RDF) XML attributes (as properties) cannot be inherited by its descendents.
• There are no classes in XML.
• The grouping of parent-child elements in XML is an non-transitive hierarchy that does not allow inference.
• Querying XML for the relationships of these nodes is done with XQuery, which requires explicit path referencing.
• Transitive relationships cannot be inferred with XQuery, unlike SPARQL.
• The sequences that represent a text are not defined by the XML, and still must be inferred by the reader.
• Dereferencing an element in the XML from its context requires either an xpath or xpointer fragment designation.
• TEI-XML cannot express everything that needs to be said about an element. It has severe spatial and semantic constraints.

RDF Structured Text Concept:

• An RDF structured text is a list of lists (where a list is defined as a ordered sequence) with the character being its base unit.
• A word is a list of characters, a line is a list of words, a division is a list of lines, a chapter is a list of divisions.
• A page unit is defined by a list of lines first, because the delimitation of a page is not necessarily constrained by the division.
• The line is also a key externally referenceable resource

A JSON-LD representation of a text page could look like this:

{
  "@context": "http://structured-text.org/context.json",
  "@id": "http://example.org/fcrepo/rest/edition/0003/page/p001",
  "@type": "st:TextPage",
  "label": "Seite 1",
  "lines": [
       {
            "@id": "http://example.org/fcrepo/rest/edition/0003/line/ln00001",
            "@type": "st:TextLine"         
       },
       {
           "@id": "http://example.org/fcrepo/rest/edition/0003/line/ln00002",
           "@type": "st:TextLine"
       }      
   ]
}   
   

The representation of a line in RDF:

<http://localhost:8080/fcrepo/rest/edition/00027/line/ln001> <http://structured-text.org#hasWords> _:c002 .  
_:c002 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://localhost:8080/fcrepo/rest/edition/00027/word/w000001> .  
_:c002 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:c005 .  
_:c005 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://localhost:8080/fcrepo/rest/edition/00027/word/w000002> .  
_:c005 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:c008 .

• and so on... until the last statement which would look like this:

_:c0050 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://localhost:8080/fcrepo/rest/edition/00027/word/w00020> .  
_:c0050 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest><http://www.w3.org/1999/02/22-rdf-syntax-ns#nil>

Representation of a division in RDF:

• Constructing a division is the same as the page, except it would look like this:

<http://localhost:8080/fcrepo/rest/edition/00027/div/d001> <http://structured-text.org#hasLines> _:d002 .  
_:d002 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://localhost:8080/fcrepo/rest/edition/00027/line/ln001> .  
_:d002 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:d005 .  
_:d005 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://localhost:8080/fcrepo/rest/edition/00027/line/ln002> .  
_:d005 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> <http://www.w3.org/1999/02/22-rdf-syntax-ns#nil>

About Blank Nodes

• The use of blank nodes is integral to the concept of list sequences in RDF,
• The JSON-LD representation of a RDF list is an array that when normalized, must include these blank nodes.
• This is the only mechanism where the resource can be dereferenced from the list structure and still retain the order of the list.
• In Fedora, when a blank node is inserted, an uuid is generated for it and it is put in the '.well-known' container.
  • example: .well-known/genid/ae/22/a3/25/ae22a325-d89f-4948-8de7-4f698e44c844

Every word is an RDF resource!

• Infinite space for statements
• Expanding on the word resource, for example:

<http://localhost:8080/fcrepo/rest/edition/00027/word/w000001> rdfs:label "In"@de .  
<http://localhost:8080/fcrepo/rest/edition/00027/word/w000001> tei:rendition tei:center 

etc...

• The enrichment of TEI annotations is retained as statements about the resource.

About Word groups

• a word group (or phrase) can also be represented as resource
• for example, "J.F. Blumenbach" is two words that represent one resource entity.
• The construction of the word group is the same as the line.

Why Construct Texts as RDF?

• The objective is to integrate the presentation of digital resources with the presentation of the annotation list metadata.
• The digital resources are stored in Fedora.
• The resources in an annotation list typically include a transcription of the text.
• Indexing this text in association with the presentation object is important for discovery.
  • If the text is kept as TEI, the TEI needs to explicitly reference the image resource (with a hard link). And, the indexing of the TEI has to be done separately. (Not optimal!)
• enabling powerful SPARQL queries to create dynamic resource presentations based on the annotation encoding criteria is a possibility.