WP 3: Enhancing eLearning with semantic knowledge

1
WP 3 Enhancing eLearning with semantic knowledge

• Kiril Simov
• Zürich, 25.01.2008

2
Plan of the Talk

• Preliminaries the results of the first year
• Ontology to text relation
• Evolution of LT4eL ontology
• Comparison of text search and ontology search
• Conclusions

3
WP3 Goals

• Creation of ontology and ontology system to
  support
• Classification of learning objects
• Annotation of LOs with concepts ontology search
• Multilingual search for learning objects
• Media between the different languages

4
Where We Stand First Year

• Target Domain Computer Science for non-computer
  scientists
• The creation of ontology is data-driven (based
  on LOs in all languages) 707 domain concepts,
  306 OWN concepts, 221 DOLCE
• The domain concepts reflected mainly the keywords
  in the domain, which were annotated by all
  partners

5
Where We Stand Second Year

• Formal evaluation of the ontology
• Consistency OntoClean Methodology (Guarino)
• Coverage of the domain Annotation of the LOs in
  all languages new concepts
• The shape of the ontology Introduction of
  grouping concepts (916 domain concepts)
• Better search support Comparison between text
  search and ontology search
• Integration within ILIAS
• User validation of the new functionalities

6
Annotation of LOs

• Annotation of the text with concepts
• Identification of the text chunk that will be
  annotated
• Assigning of all possible concepts for the chunk
• Concept disambiguation
• Model of the relation between the ontology and
  the text

7
Mapping Lexical Varieties
8
Example from the Dutch Lexicon

• ltentry id"id60"gt
• ltowlClass rdfabout"lt4elBarWithButtons"gt
• ltrdfssubClassOfgt
• ltowlClass rdfabout"lt4elWindow"/gt
• lt/rdfssubClassOfgt
• lt/owlClassgt
• ltdefgtA horizontal or vertical bar as a part
  of a window,
• that contains buttons, icons.lt/defgt
• lttermg lang"nl"gt
• ltterm shead"1"gtwerkbalklt/termgt
• lttermgtbalklt/termgt
• ltterm type"nonlex"gtbalk met
  knoppenlt/termgt
• lttermgtmenubalklt/termgt
• lt/termggt
• lt/entrygt

9
The Relations between the Lexicon, the Grammar
and the Text
10
Stages of Semantic Annotation

• Preparation for the semantic annotation (grammar,
  layouts, DTD, constraints)
• Actual annotation semi-automatic, involves
  annotator choice points
• The tool used for the task is CLaRK System

11
The Role of the Grammar and Constraints

• Grammar assigns all possible concepts per text
  segment
• Constraints two variants
• Constraint 1 (Select Concept)
• stops at each annotated node
• introduces artificial ambiguities EXTEND, and
  ERASE
• Example Internet vs. Wireless Internet
• Constraint 2 (Select LT4eL Concept) at later
  stage, when the grammar shows a better performance

12
Interaction Among Modules
13
Evolution of LT4eL Ontology

• Identification of new concepts during the LOs
  annotation and Validation Phase I
• Change of the upper ontology to DOLCE Ultralite
  a simplified version of DOLCE
• Addition of relations better navigation over
  the ontology and understanding of the concepts

14
Formal Evaluation of Ontology Search

• Search for paragraphs with query formed on the
  basis of Concepts from ontology
• Search for paragraphs with query formed on the
  basis of Terms in the lexicons
• Cases
• Ambiguous term depends on WSD - help
• Unambiguous specific terms slide and
  presentation
• General terms program
  (ontology inference)

15
Search without Inference
16
Search with Inference
17
Results

• Precision RetrievedRelevant/AllRetrieved
• Recall RetrievedRelevant/AllRelevant
• F (2RecallPrecision)/(Recall Precision)
• Assumption (CLEF) All relevant paragraphs are
  among the paragraphs retrieved by the two kinds
  of search (not 100 correct, but good in practice
  for comparing the searches)

18
Procedure

• Form two queries ontology and terms
• Extract two sets of paragraphs
• Mix the two sets and delete the repetitions
• Assess the relevance of each paragraph
• Distribute the relevance back to the two sets
• Calculate the measures

19
Results
20
Conclusion

• We have selected a methodology for the evaluation
  of LT4eL ontology
• We have defined a model for ontology to text
  relation
• We have annotated the LOs with concepts
• We have enriched the ontology in order to be
  better with respect to the tasks of the project

21
Plans for the Last 6 Months

• We have to enrich the ontology
• We have to construct lexicons for the new
  concepts (including the upper ontology)
• We have to create grammars and annotate the new
  concepts in the LOs
• We have to use the new resources in the next
  Validation Phase