Exploitation of Dynamic Information Relations in the Service-Oriented AFRL Information Management Systems Andrzej Uszok, Larry Bunch, Jeffrey M. Bradshaw.

Exploitation of Dynamic Information Relations in the Service-Oriented AFRL

Information Management Systems

Andrzej Uszok, Larry Bunch, Jeffrey M. Bradshaw

James Hanna, Albert Frantz

Outline

AFRL Information Management System and its limitations

Motivation for Dynamic Information Semantic model of information Document ontology and annotation Information relevance ontology and folksonomy Architecture of IMS extensions with Dynamic

Information Conclusion

AFRL Information Management System

Information Management System consists of a server, a client interface (CAPI), and associated clients

IM Server consists of set of services performing information brokering and dissemination

Information is packaged into Managed Information Objects (MIOs)

– Consists of metadata (XML) and payload (binary)

Client operations include– Publish– Subscribe– Query

Information Space Overview

During the previous KSCO we presented Federation Service for IMS.

Information Space

Information Catalogue

Stored Info. Index

Information Models

Client Status

Consumers Producers

Information Repository

Transient Info

Persistent Info

Archived Info

CAPI

CAPI

LocalInformation Space

Producer Producer

Consumer Consumer Consumer

Publisher Data(Access Control,

Information Preprocessing,QoS Enforcement,

Information Transformation)

Publisher Channel(Registration,

Advertisement,Feedback, RFIs)

Broker(Consumer Data)

Consumer Control (Registration, Subscription,

Search and Query Requests)

Quality of ServiceManagement

IMS Limitation

Require applications to map to “single managed information format

Information instances immutable Subscription and Query scoped by information type Lack of support for unstructured query Limited usefulness and adaptability of IMS to the coalition

information sharing.– Strict restriction on used types and formats of information– Inability to selectively extract parts of the information intended for

sharing

Current Assumptions of IMS

The type of information object used for publication, subscription and query has to be registered in the Information Type Service.

Vision for Dynamic Information

Flexible information model supporting a variety of existing information schema with rich semantic

Ability to correlate dynamic information in order to provide comprehensive mission information

– a generic semantic relationship representation – discovery of relationships among semantically-related

information – dynamic information groups

Use of semantic relationships to provide– Support subscription and Query across multiple information

types – Unstructured Query, system creates new information "on the fly”– Stream annotations

Syntactic Mapping

Requirement to support variety of information type representation: XMLSchema, DDL, RDF, Office etc.

Mapping from less expressive representation to more expressive is feasible

Common expressive representation eliminates needs for one to one mapping

Relations are easily represented in OWL/SPARQL RDF/OWL representation provides an explicit semantics

allowing for establishment of information relationships The resulted mapping is annotated with information

of the origin as the reverse translation is necessary

Semantic Linkage Technology

Existing standards and techniques for semantically linked data

Uniform semantic layer provides marriages across various standards allowing automated pattern searches and queries across previously unconnected sources.

Document metadata: author, create date, etc. Document elements: titles, text runs, captions,

etc. Shapes and picture elements Structure

– Nested documents– Documents parts

Document Formats

Ontology for Documents

Fragment of Document Ontology

Capabilities– Create an OWL description for the contents of a

MS Powerpoint and Wordx document– Create an image for each slide in a Powerpoint

Technologies– Off-the-shelf parser

Apache POI v.3.7 http://poi.apache.org

– KAoS Java libraries for constructing OWL Relies on Jena http://jena.sourceforge.net

Office Mapping to Ontology

Natural Language Indexing

Published information is processed by GATE text processing system

– Named-entity-extraction – who, when, where and what– Pattern matching grammar rules are based on the ontology

classes Uses UCORE-SL (extended)

Annotated Information is indexed by Apache Lucerne Semantic annotation and metadata are stored in

ontology store (Jena TDB) Support for free text search integration with SPARQL

UCORE-SL

Military ontology associated with UCORE XML message format standard

Analysis of Information Relevance

Support for Folksonomy

Free tagging are needed in many situations:– There can be omissions in ontology– Concepts for new things and phenomena – Instance data, e.g., persons, places, events etc. too

numerous Integration of new free keywords into

ontologies in an annotation environment– The keyword are mapped to the existing ontology

usually through the rdfs:subClassOf property

Service Oriented IMS

Latest AFRL reincarnation of IMS Phoenix has Service Ordinated Architecture

Set of independent, flexibly deployable IM services– Submission, Subscription, Information Brokering, Dissemination,

Repository, Query, Type Management, Event Notification, Service Brokering, Session Management, Information Discovery, Security, Stream Service

This features allowed us to:– gradually extended selected services of the IMS with the new

capabilities – dynamically configured IMS

Dynamic Information Type Management Service

Parser specific to the given type provides a uniform view for the relation discovery mechanism (additional schema annotations)

Precompute (or acquire from the user) information relationships when new information types are created in the repository

The precomputed relationships are the potential relationships realized by information depending on their actual values

GUI for New Information Type Service

Create and manage a unified model of information types Import new type descriptions from existing XML

Schema, DDL and Office documents. These types are mapped and integrated into the unified OWL model

View, Update, and Define– information type descriptions in OWL– relationships among type descriptions– contexts in which certain information type relationships hold– combination definition for a resulting product

Type ManagerType Manager

Types

Air Tasking Order

ID: ATO

Relations for Selected Type+ -

Creator: [email protected]

Weather Forecast

ID: WeatherForecastCreator: [email protected]

Original: schema/ATO.xsd

Part OfATO

Creator: [email protected]:

SITREP

Creator: [email protected]

Context: time within 10 minuteslocation within 1 km

Weather Report

ID: WeatherReportCreator: [email protected]

Original: schema/WForecast.xsd

Original: schema/WReport.xsd

http://afrl.af.mil/Model.owl#ATO

Subclass OfPeriodicReport

Creator: [email protected]

Contributes To

WeatherForecast

Creator: [email protected]: time within 30 minutes

location within coverageArea

+ -

Subject Search < > ResetObjectRelation Go

Dynamic Information Type Management GUI

List typesin the typerepository

List typesmatching

search criteria

Subject-Relation-Objecttype search

Hyperlink to OWL(text or graph)

Hyperlink to original type definition file (XSD,OWL,DDL)

Partial text match of type ID’s

Default values indicate open-ended search, e.g.

o weather typeso types with any relation to weathero types that contribute to a weather forecast

Hyperlink to type list

Select a type to view its relationsAdd a

new type inType Editor

Edit TypeEdit Type

TB

FL

TM

OV

UA

Original Property

Xsd:String

xsd:Integer

xsd:datetime

xsd:latlon

xsd:UUV

Original Range

turbulence

altitude

datetime

location

urgency

OWL Property

model:Turbulance

xsd:Integer

time:Instant

model:LatLon

model:UrgencyLevel

OWL Range

Weather ReportLabel

URI

External Type

http://afrl.af.mil/Model.owl#WeatherReport

Description of current weather conditionsDescription

XML Schema

0+

1

1

1

1

Cardinality

Property Mapping

2011-01-01 13:55:01.1

Date

[email protected]

User

Change History

Initial version

Description

http://afrl.af.mil/Model.owl#PeriodicReporthttp://afrl.af.mil/Model.owl#PointLocatedEntity

Superclasses+ -

Source

Browse

schema/Wreport.xsd

View

urgency

location

datetime

altitude

turbulence

External Type

Source

Database Table

AFRL Weather DB

Table FORECAST

Type Editor

Human-readable label

Unique type identifier (RDF)

Original type description

Human-readable description

Parent classes from the unified model

Mapping of original propertiesto OWL

Browse for an XSD, OWL, DDL file or

Office

Original properties are parsed from the source file

OWL properties

are editable

Edit Relation for WeatherReportEdit Relation for WeatherReport

Relation Type Contributes To

time of weatherReport within 30 minutes before startTime of weatherForecast andlocation of weatherReport within coverageArea of weatherForecast andproperty of weatherReport relates to property of weatherForecast +

Relation Context Expression

Related Class WeatherForecast

Browse Classes...

Relationship Editor

Subject class of the relationship

Object class of the relationship

Type of relationship

Context in which the relationshipapplies, expressed using the

SPARQL syntax for triple patterns and FILTER statements

Click forClass

Browser

Click for Date and Time

Editor

Click for Area Editor

SPARQL expression.created using hypertext

interface

Dynamic Publication

When a publication is made, additional publications are generated based on relationships to other types.The information itself is also extended with additional related available information from the persistence information repository. Information groups are precomputed, based on existing subscriptions, with related persistent information facilitating semantic dissemination.

Dynamic Subscription and Query

• LARQ - Free Text Indexing for SPARQL; ability to perform free text searches combined with structured search

• Relation based Result enhancement – Searches for related information have their expression generated based on the original expression and the values in the current result under consideration

Combination of semantic with natural language and human annotation of documents provides rich space for discovering relations between information.

Flexible model for information allow for greater adaptability of IMS for coalition information sharing

Performance demand for the new semantic information mechanisms need to be controlled by policies and addressed by more computing power – cloud computing

Conclusion