2. ? Searching the WWW today document retrieval keyword based search user
document(s)
query
result
search engine
text
images
video
audio … ?
Searching today:
Query consits out of (some) keywords [sometimes connected by boolean operators)
Search engine provides list of ‚related‘ documents (related = containment of keyword)
How to connect keywords to documents?  search engine index
Documents are analysed by information retrieval techniques (data mining)
Keywords are extracted from text (word stemming, relevance weight, etc…)
PROBLEMS:
Language inherent polysemy (ambiguity, synonyms, …)
Missing keyword context
Document context not extracted by data mining
Semantic web relies on metadata [i.e. data that describes documents in machine accessible format]
Metadata solves disambiguation problems
Metadata might solve context problems
But….from where to get the metadata??
keywords
+ metadata
documents to read
search engine index
document content

3. There is already (often unused) Metadata
semantic annotation
document
index
TOC
references
Document contains inherent semantic annotation
Index  internal conceptual knowledge
TOC  internal structural knowledge
References  referential knowledge
index (conceptual knowledge)
TOC (structural knowledge)
references (referential knowledge)
basis of semantic document annotation

4. Documents, Tags, and Annotations
<b> Lorem</b> ipsum dolor sit amet, <br/>consectetuer adipiscing elit. <br/> <a href=“……“ title=“..“/> Sed orci purus, semper eget, <br/> tristique quis, adipiscing <br/> <!--<rdf:annotation user=“…“ tag=“…“…/> posuere, erat. Aenean <br/> ultricies odio id sem. Sed <br/><h1> nec felis sit amet ante </h1> tempor sagittis. Vestibulum <br/> est nunc, lobortis cursus, <br/> semper vel, pulvinar sed, <br/> odio. Vestibulum blandit…
document
consists of
annotations 
strings
associate distinguished document parts with metadata
Documents might carry inherent semantic annotation (tags)
 for metadata association
 for user/creator association
Metadata can be provided by author [ document inherent] or by user [ document addendum, see Web 2.0]
smallest addressable document unit

5. Documents, Tags, and Annotations
Examples
smallest document unit: word
higher order units: sentence, paragraph, page, chapter, part, …
book
Different kind of contextes
Syntactic structural: physical units
Semantic structural: logical units
smallest document unit: pixel
higher order units: blocks, macro blocks, slices, frames, objects, scenes, acts,…
video

6. Logical Document Structure
Table of Contents (TOC) from structural tags
page 1
page 2
page 3
page 4
page 5
page 6
page 7
page 8
page 9
page 10
page 11
page 13
page 12
page 14
page 15
1.1 Introduction
1.2 Definition of the basic formalism
1.3 Reasoning Algorithms
2.1 Introduction
1.1 OR-Branching finding a model
1. Basic Description Logics
2. Complexity of Reasoning
can be specified
explicitely (structural information) or
implicitely (formatting information)
can be associated with names/titles
can be used for document navigation
Also for video:
Automatic scene detection (start/end point)
Automatic segmentation of video
Annotate segments manually / or with multimedia mining techniques
Encode Table of Contents (TOC) with MPEG-7 encoding
Basic Description Logics 1
Introduction 1
Definition of the basic formalism 5
Reasoning algirithms 7
Complexity of Reasoning 11
Introduction 11
OR-Branching: finding a model 12

7. Conceptual Document Structure
root
SUB
SUB
rodent
field mouse
conceptual structure
SEA
SUB
SUB
SUB
SUB
SUB
SUB
hamster
dentision
beaver
habitat
meadow vole
prairie vole
SUB
SUB
Using explicitely given conceptual document structure together with logical document structure to define the document index
SUB  subsumption
SEA  see also
Can be considered as a kind of ontological skeleton
Covers concepts of the document and their relationships
Using implicitely given conceptual structure requires understanding of document content
Explicitely given conceptual structure (only a small fraction of entire conceptual structure) can be defined by
document author (e.g., index entries, external metadata)
document users (e.g., social tagging)
The conceptual document structure can also be used for document navigation
incisor
rotation of teeth 1 2 3 4 5 6 7 8 9 10 11 13 12 14 15 16
Paragraph1.1
Paragraph1.2
Paragraph1.3
logical structure

8. Conceptual Document Structure
rodent, 1
beaver, 10, 11
dentision
incisor, 4
rotation of teeth, 5
hamster, 2 - 4
see also meadow vole …
field mouse, 13, 15
prairie vole, 16
meadow vole, 16
habitat, 15
see also rodent
Conceptual Structure also comprises explicit links to complex Ontologies that describe the meaning of the documents content
With simple data mining techniques, a fraction of the conceptual structure of multimedia docuzments (e.g. MPEG videos streams, lecture videos) can also be made explicit  the other talk about automated annotation
Document index (MPEG Index) can be encoded as MPEG7 File and can also be used for navigation
Document Index

9. Referential Document Structure
Internal links: References between parts of the same document e.g., see / see also, footnotes, figures, comments…
External links: References between different documents e.g., bibliographic references and citations,…
Only a fraction of the entire referential document structure is given explicitely
Graph Visualization (Link Graph)
together with logical document structure  table of figure, references, …
Implicitely given referential structure  requires text understanding capabilities

10. The Structures in Concert
root
SUB
SUB
conceptual structure
rodent
field mouse
SUB
SUB
SUB
SUB
SUB
SUB
hamster
dentision
beaver
habitat
meadow vole
prairie vole
SUB
SUB
Given, you want to know something about rodents……..what do I have to read (in 1 document /book or even within a collection of documents)
I want to read as few as possible, but I want to understand everything as good as possible
incisor
rotation of teeth
referential structure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
logical structure
Paragraph1.1
Paragraph1.2
Paragraph1.3

11. Thank your for your attention!
Conclusion
Documents have intrinsic logical, conceptual and referential characteristics
There are complex dependencies among the document structures carrying those characteristics
Logical, conceptual, and referential structures along with their interdependencies should be made explicit ( meta data)
Applications should maintain and use those meta data, e.g. for
authoring
navigation
searching
Thank your for your attention!

12. The Structures in Concert
All three structures in concert can be used for
Document reading tours (extended document retrieval)
goal oriented selections of documents (what is mandatory to understand the topic under consideration?)
with additional reading directions (which document unit to read in what order)
by also considering user annotations, personalized reading tours can be suggested (dependent on prior knowledge of the user)
Collaborative authoring (avoiding ambiguities or duplicates, support index generation and cross referencing,…)
Compute answers… (with the help of sophisticated reasoning and additional means of data mining and content understanding)

13. Related Work  Topic Maps
Topic Maps represent concepts and relationships (conceptional structure and (part of) referential structure)
Topic Map
rodent
part of 1
type
association type
role
whole
part 10
topic
beaver
dentision
Topic Maps do not include the logical document structure 11
association
Resources