Introducing Knowledge Access in SEKT

1
Introducing Knowledge Access in SEKT

• Alistair Duke, BT

2
What is Knowledge Access?

• Delivery of knowledge to end user
• A tool providing access to one or more knowledge
  bases
• E.g.
• Search Interface
• Knowledge Sharing system
• Summariser
• Visualisation

3
Knowledge Access Problems

• Information Overload
• Search Engines
• High recall low precision
• Syntactic search is imprecise
• High cognitive load
• Not contextually aware

4
Knowledge Access in SEKT
Knowledge Access Tools
Human Language Technology
Knowledge Discovery
Ontology Metadata Mgmt
5
Knowledge Access in SEKT

• Context-aware tools for access to
  semantically-annotated knowledge
• search, browse, visualise, summarise, share,
  infer
• integrated into day-to-day business processes
• automatic knowledge delivery based on current
  context (activity, location, interests)
• support multiple end-user devices

6
Knowledge Access Tools

• User Profile Construction

7
Knowledge Access Tools

• Search Browse

8
Knowledge Access Tools

• Knowledge Sharing

9
Knowledge Access Tools

• Visualisation

10
Knowledge Access Tools

• Knowledge Generation

11
Knowledge Access Tools

• Device Independence

12
User Profile Construction

• A Web-based user profile is automatically
  generated while the user is browsing the Web.
• It is represented in the form of a
  user-interest-hierarchy
• The root node holds the users general interest,
  while leaves hold more specific interests
• Nodes of current interest are determined by
  comparing nodes with the topics computed from
  recently visited pages.
• The user profile is visualized on the SEKTbar
  (Internet Explorer Toolbar)
• The user can select a node in the hierarchy to
  see its specific keywords and associated pages

13
Demonstration
14
Search Browse

• Searching both semantic metadata and free text
• Two Prototypes
• SEKTagent
• Periodic queries based on named entities
• Semantic Search Engine and Interface
• Building a search interface directly from the
  ontology
• Uses BT Digital Library data as use case

15
SEKTagent

• Provides automated knowledge delivery based on
  search queries augmented with semantic info from
  ontology
• Periodic agent search for named entities
• E.g. a person in an organisation
• Result-led indexing
• Adds relevant pages to start points for next
  crawl.

16
(No Transcript)
17
Search Engine Interface

• Combined full-text and semantic search.
• Providing a highly configurable front-end to
  present a logical search interface
• Moving towards presenting knowledge rather than a
  list of web pages

18
(No Transcript)
19
Knowledge Sharing

• Sharing Knowledge amongst a community of interest
• Distance between semantic profiles can be used to
  determine
• what to share
• who to share it with
• System should adapt to changing communities and
  feedback from users

20
Visualisation

• Systems exist to allow ontologies to be shown
  graphically e.g. Jambalaya
• Practical ontologies do not lend themselves to
  direct visualisation
• In SEKT we introduce a visualisation ontology
  layer
• General purpose Visualization
• Input Ontology
• Visualization Ontology
• User Rules
• Default Rules
• Interactive
• Browse through concepts/instances
• Interactive Objects
• Profile-based

21
Visualisation - Example

• Types of Visualisation

Graph
Hall
Viz Ontology
Domain Ontology
Mappings
Publication_Entity ? Authors ? Published
Works Publication_Entity ? Works ? Books
Entity ? Agent ? Person Entity ? Agent ?
Literary Work Relation ? Responsibility Relation
? Creation Relation
22
Knowledge Generation

• Generation of natural language from ontological
  information
• Generated summary can be in HTML or text
• Length restrictions can be imposed
• Module is a GATE component which allows the user
  to customise their parameters
• Profile aware

23
Demonstration
24
(No Transcript)
25
(No Transcript)
26
Knowledge Repurposing

• DIWAF - Device Independence Web Application
  Framework
• Delivers knowledge to a user in a format
  appropriate to the device to which they currently
  have access
• Recognises device from HTTP request (Uses RDF
  device ontology for device recognition)
• Separation of data from presentation
• Aim is to provide device independence for SEKT
  applications
• Extend for context awareness based on user /
  geographic profile

27
Knowledge Repurposing

• Requests are intercepted by server
• Selection of template based on device
  characteristics (specified by author)
• Data can be read from SOAP service, XHTML, SQL
  database, etc.

28
Demonstration
29
(No Transcript)
30
Integration / Applications

• Tools will be integrated in SEKT via SIP (SEKT
  Integration Platform)

Text Documents
GATE
Search Indexer
Search UI
KAON
31
Summary

• Knowledge Access is about providing knowledge to
  the user at the correct time in the correct
  format
• In SEKT it links the key technologies to the end
  users
• The Knowledge Access tools are being trialled on
  industry case studies

32
Introducing Knowledge Access in SEKT

• Alistair Duke, BT
• alistair.duke_at_bt.com