Shared Ontology for Knowledge Management

1
Shared Ontology for Knowledge Management

• Atanas Kiryakov, Borislav Popov, Ilian Kitchukov,
  and Krasimir Angelov

Meher Shaikh
2
Overview

• The authors present an approach for semantic
  searching on the web.
• Indexing schema based on entity occurrence.
• Demonstrate scalable implementation of the
  indexing.
• KIM platform allows semantic annotation,
  indexing of documents with respect to named
  entities (NE).
• CORE module based on co-occurrence of entities.
• User Interface CORE Search and Timelines

3
Objective

• The contemporary search engines uses ranking to
  provide the relevant information based on string
  tokens.
• Involves semantic analysis of the data on the
  web.
• Example query telecom company in Europe John
  Smith director
• Information need A telecom company in Europe, a
  person called John Smith, and a management
  position.
• A document containing the following sentence
  would not be returned using conventional search
  techniques.
• The search engine needs to be able to consider
  several semantic relation and inference rule to
  return this above document.

4
Traditional IR

• Vector-Space Model (VSM) The documents are
  characterized by the token appearing in them. The
  model evaluates the similarity between the query
  tokens and the tokens appearing in documents to
  retrieve and rank the documents.

5
Shared Ontology Approach

• Combines the advantages of the semantic
  repository and the raw power of relational
  databases.
• Semantic repository allows inferring and querying
  on top of formal knowledge. The relational
  databases can handle large volumes of data
  efficiently.

6
KIM

• The platform provides infrastructure for
  automatically extracting named entities (semantic
  annotations) from the unstructured text. This
  includes attributes and relations. The extracted
  information is presented in a knowledge base
  called the semantic repository. The semantic
  annotations are then used for indexing of the
  documents.

7
Semantic repository repository of entities and
their relations
Example The semantic repository infers that
London is part of UK
8
Partial Architecture of KIM Platform
9
CORE module

• Extension of KIM platform with advanced UI
• Based on robust open source platforms specialized
  in ontology management, text mining and IR.
• Focuses on co-occurrence of entities.

10
CORE module cont..

• Maintains bi-directional relations between entity
  and documents.
• This allows retrieval of entities by documents in
  addition to retrieval of documents by entities.
• Provides incremental searching, ranking, and
  tracking and popularity timelines of these
  entities.

11
User Interface CORE Search
12
User Interface Timelines

• Timelines interface allows trends to be
  calculated and conveniently viewed and
• navigated through.

13
(No Transcript)
14
Conclusion

• More Meaningful content extracted using semantic
  analysis and inference.
• The KIM Platform and its CORE module currently
  achieve real-time retrieval from about a million
  documents and a million of entity descriptions.
  Work is ongoing to deal with the large amount of
  web resources
• Synchronization techniques between database and
  the semantic repository yet to mature.