from knowledge bases to knowledge infrastructures for intelligent systems

32
From Knowledge Bases to Knowledge Infrastructures for Intelligent Systems Mathieu d’Aquin Professor of Informatics, Insight Centre, NUI Galway, Ireland @mdaquin - mdaquin.net

Upload: mathieu-daquin

Post on 21-Jan-2018

299 views

Category:

Technology


0 download

TRANSCRIPT

From Knowledge Bases to Knowledge Infrastructures for Intelligent Systems

Mathieu d’AquinProfessor of Informatics, Insight Centre, NUI Galway, Ireland@mdaquin - mdaquin.net

Old-School Knowledge-Based Systems

Cancer treatment guidelines

Formalised Knowledge

Base

Interactive Decision Support

Knowledge encoding and representation

Reasoning and interrogation

Scale….

Data

Information

Know-ledge

Machines

Scale…

Data

Information

Know-ledge

Machinesstorage

interrogation

processing and analysis

reasoning and decision

The Web of Data

The Web of Data

The Web of Data

Gene Ontology

FMA OntologyLODE

BIBO

Geo Ontology

DBPedia Ontology

Dublin Core

FOAF

DOAP

SIOC

Music Ontology

Media Ontology

rNews

Need a different kind of intelligent systems

Watson Semantic Web Search Engine

http://watson.kmi.open.ac.uk/

Watson Semantic Web Search Engine

Accessing ontologies on the semantic web through smart APIs - making it possible to build intelligent systems using online ontologies as their knowledge bases.

Example application - search

Example application - ontology edition

Example application - ontology matching

Example application - question answering

Shameless Plug...

Scale…

Data

Information

Know-ledge

Machinesstorage

interrogation

processing and analysis

reasoning and decision

Watson: An attempt to scale up the knowledge level

Next step

Data

Information

Know-ledge

Machinesstorage

interrogation

processing and analysis

reasoning and decisionUsing the knowledge level...

To make large scaleinformation and datalevels more exploitable

Example: The MK Data Hub

Data Infrastructure for the city of Milton Keynes, enabling sharing and consuming varied and diverse city scale data.

Example applications

Example: MK Insight

But… a large number of datasets for a large number of applications

MK Data Hub

Analytics

Integration

Curation

Storage

Import

Sensor Data

Local Stats

Gov. Open Data

...

Mobile Apps

Dashboards

Business Intelligence

Social Web Apps

...

Data cataloging needs to do more...

Data cataloging component to index data based on their provenance, categories, format, existing use, etc.

But needs to do more to answer questions such as :

- Can I use those data for a commercial application? Do I need to attribute somebody? Even after processing?

- What can this data do? What kind of things I can apply on it?

Ontological approach to data policies

Explicit, semantic representation of the licences attached to data

As well as the data flows through which they are processed.

Automatic propagation of policies through dataflows

Understanding what data can answer

Example of using formal concept analysis to extract relevant questions from an RDF (graph) dataset.

Ongoing work on generating interactive interface to ontology-based data

Service code

Area

Restaurant

Organisation

isa

population

deprivation

locatedIn

rating

employee

Person The population of Walnut Tree is 4096

What is the population of Walnut Tree?

Towards populating ontologies based on dialog

Mood

Good Mood

Bad Mood

Very Bad Mood

Very Good Mood

isa

isa isa

isa

excellent

type

horrible

typebad

OK

good

typetype

type

betterworseinverseOf better

TT

better

better

better

Thanks! I don’t know “great”, is it better or worse than “OK”? ...

Alexa, tell moody that I’m feeling great!

Towards the automatic exploitation of dataExample, in autonomous agents, using an ontology that provides a typology of datasets and of data analytics techniques, making them better able to automatically exploit the data they come across.

Conclusion

Knowledge representation and ontology engineering have gone a long way from top down, closed, domain centric knowledge-based systems.

From encoding expert knowledge to dealing with scale, variety and diversity.

Now, becoming central in the necessary automation of information processing, making data analytics and mining more directly accessible, with fewer bottlenecks.

Contact: @mdaquin - mdaquin.net