Semantic Analysis of Text

1
Semantic Analysis of Text

• Michal Laclavík
• Ústav Informatiky SAV

2
Outline

• Motivation
• Graph and Object representations
• Ontologies Organizational Memories
• Context Content
• Approaches and Technologies
• Semantic Annotation
• Vision of Semantic Web
• Vision of Semantic Organization

3
Motivation

• Documents are valuable source of information
  knowledge in eGov and also in eBusiness,
  eCommerce
• Intranet, Internet systems HTML documents
• Electronic communication organizational task
  and processes text
• Content Context

4
Structured Information

• Text not structured not understandable for
  computers
• Goal to provide structured information
• Based on structured information it is possible to
  provide knowledge
• Types of structured knowledge
• XML based versus others (Object database, RDBM,
  logic based, rule-based)
• RDF
• OWL ontologies, OWL-DL

5
Organizational Memories

• Metadata about Resources
• Documents
• People
• Customers
• .
• Based on Ontologies Semantic Web Technologies
• Tied with Knowledge Management Solution

6
Context

• We need to detect context of information
  knowledge in order to provide it in user specific
  context
• When, by who, why was document created
• In what activity, what process, what category
• Not only text of Documents is relevant !
• Should be tied with user context analysis
• Workflow, email, file system - computerized
  activities
• Context detection - from electronic communication
  channels analysis

7
Content

• We need to detect content of document
• Category
• What document talks about
• Semantic of the document
• Analysis of Text
• Finding objects from problem domain in text

8
Existing Solutions

• Mainly stamping HTML documents for reader
• Annotea, Ruby annotation,
• Often creating own structured XML data or new
  ontology
• Specialized Browsers
• Magpie
• Manual Annotation Tools
• CRAM

9
Manual Annotation Browsing
10
Semi-Automatic Annotation

• Natural Language Processing Training sets
  Learning
• Neural Networks, Genetic Algorithms, Hidden
  Markow Models, ..
• Document Structure Based
• useful for HTML
• Wrappers
• Learning of Structure patterns
• Pattern Recognition
• C-PANKOW
• Ontea

11
C-PANKOW

• Pattern matching
• Google API request
• Creating ontology individual of most typical
  class (type)

Is a pattern for Niger by Google API
12
C-PANKOW

• QTag
• Detection of Nouns
• plural proper noun (NPS)
• plural common noun (NNS)
• common noun (NN)
• proper noun (NP)
• adjective (JJ)
• interjection UH

13
Ontology based Text Annotation OnTeADeveloped
at IISAS
14
Motivation Objectives

• Detecting Meta data from Text
• Preparing improved structured data for later
  computer processing
• Structured data are based on application ontology
  model

gt
15
Domain Ontology Job Offers
16
Ontea Architecture

• Regular expressions applied to find individuals
  corresponding with text
• Text Class in Ontology gt New individual is
  created output individual
• Output individual can have properties of certain
  types
• Individuals found in text are properly assigned
  as properties of result individual
• Inference is used

17
Simple Pattern Ontology

• Can be used for any Application
• Can be extended

18
Results - Jobs

• Creation of Jobs Metadata
• Same structure for each document
• Can be used for processing, presentation

19
Success rate
2
1
3

• 500 documents
• Found more elements

20
Conclusion on Annotation

• text gt representation of the text in the domain
  ontology model
• Used in web job offer documents application and
  Flood pred. application
• We are applying it in email analysis app

21
Vision of Semantic Web

• The Semantic Web is a mesh of information linked
  up in such a way as to be easily processable by
  machines, on a global scale. You can think of it
  as being an efficient way of representing data on
  the World Wide Web, or as a globally linked
  database.(Source http//infomesh.net/2001/swint
  ro/ - The Semantic Web An Introduction)

22
Vision of Semantic Organization

• To have all information and data available for
  computer processing via Semantic Web technology
  (XML, RDF, OWL)
• Ontology translation not so important on one
  domain
• Document and Text analysis results using Semantic
  annotation are part of this
• Conversion or mapping of RDBMS to XML/RDF/OWL is
  another problem