automatic semantic interpretation of unstructured data for knowledge management
DESCRIPTION
The demo shows an automatic semantic analysis of Wikipedia articles about astronomy.TRANSCRIPT
Topic Maps in the Industry
TMRA 2010
Demo of an automatic semantic interpretation of unstructured data for knowledge management
Agenda
1. Demo
2. Knowledge Discovery
3. Technical Solution
Inverted approach of semantic it
1.
The demo shows twofold results of an automatic semantic analysis of Wikipedia articles to demonstrate a new approach for knowledge discovery.
Demo
1.
Analysis of Wikipedia articles about astronomy
Demo
Crawling all articles of a knowledge domain
Extracting the relevant text parts of Wikipedia pages
Extracting meta data of each Wikipedia article
Automatic semantic analysis of integrated data
on a term level to create a linked concept graph
on an object level linked data (object) graph
1.
What the demo shows
Demo
Visualization of the linked concept graph (left)
Visualization of the linked data graph (right)
Knowledge discovery by a taxonomy and linked data
Accessing information by linked data
Accessing information by derived taxonomy
2.
Isolated data becomes meaning by links to related data. Even unstructured information can be evaluated systematically by linked data and a derived taxonomy.
Knowledge Discovery
2.
Use cases for an object graph
Knowledge Discovery
Information Logistics: Relevant information will be provided automatically in the process or activity context of a user.
Portal navigation: Users can navigate according to their personal focus of interest along the dynamic links to each selected context.
Knowledge discovery: Awareness of hidden knowledge such as project synergies, sales opportunities, relevant news.
Question answering: The identification of appropriate responses, related problems, or experts on the issue.
Business intelligence: Complex queries of the object graph for reports on customer behavior, staff profiles and project analysis.
2.
Use cases for a concept graph
Knowledge discovery
Knowledge Representation: The concept graph gives an overview of key entities and facts in an unstructured data set.
Document and e-mail-clustering: Unstructured data will be grouped thematically or associated with each path in a taxonomy.
Moderated search: searches for the automatic extension of a keyword search for increased precision of the results.
Topic monitoring: Identifying new facts and new issues or topics in the news, or constellations of other publications
Taxonomy or ontology modeling and maintenance: Initial knowledge representation and identification of adaptation needs.
3.
Knowledge discovery needs a real bottom-up-approach with no initial effort on modeling a knowledge domain. The result can be exported as topic maps or combined with formalized domain knowledge of existing topic maps.
Technical Solution
3.
Implementing Content Provider
Bottom-up semantic data integration
Lean interfaces to connect any data format and source
Push and pull principle to monitor data sources
Optional bi-directional integration of data sources
Optional definition of actions for data objects in each source
Implicit data harmonization and derivation of a common model
3.
Object graph (linked data graph)
Bottom-up semantic data analysis
All relations (quadruples) are
dynamically created and updated in real-time
described by the semantic reason
weighted regarding the relevance
All relations are created by
Key attributes (syntax analysis)
Text mining (pattern analysis)
User behavior (usage analysis)
3.
Example of a graph fragment
Bottom-up semantic data analysis
3.
Concept graph
Bottom-up semantic data integration
Extraction of concepts such as names and terms in texts
Calculation of significance of extracted concepts
Identification of the co-occurrences of significant concepts
Creating a graph with significance value for nodes and edges
Dynamically updated graph caused by new data
Calculation of a hierarchical structure for a taxonomy
3.
iQser GIN Platform
ERPCRM WWW
Collaboration
Fila System
Custom Applications
Client Connector API
Content Provider API
Ana
lyze
r Tas
k A
PI
Even
t Li
sten
er A
PI
Security Layer
iQser Core
Custom Analytics / Ontologies
Custom Event Actions / Business
Logic
ESB / SOAWeb Rich-/Fat Client
Mobile
Analyzer Chain Event Processor
Objektgraph Konzeptgraph
Index
www.iqser.com
+49 172 66 800 73
Dr. Jörg Wurzer Member of the board