Building Knowledge Bases from Reusable Components Peter Clark Boeing Applied Research and Technology.

Building Knowledge Bases from Reusable Components

Peter ClarkBoeing Applied Research and Technology

Fragment of a Knowledge-Base

• “What steps are involved in bioremediation?”• “How does pollutant volume affect rate?”• “What equipment is needed?”• “What do the microbes do?”• ...

Example queries:

Bioremediation Amount Amount

Oil Fertilizer

Get Apply BreakDown

Absorb

Microbes Script

Bio-technologist

Soil Rate

environmentcontains

Q+ I- Q-I+

amount

productabsorbed

then

agent

patient agent

scriptpollutant

se

rateagent

then then

product

sesese

patient

remediatoramount

Personisa

......

...

...

...

...

agent

Potentials…

• Growth of on-line, structured information, eg– XML– On-line databases (eg. commercial, geographic)– eCommerce– NLP-generated structures

• Requirements for more than fact retrieval, eg– Search / Information Access (“best search engine wins...”)

– Knowledge Management

– NL understanding / MLT / Speech

The growing demand for knowledge processing:

The Botany KB Experience

• 10 yr effort, 20k concepts, 100k facts• Supports sophisticated question-answering

– description– prediction

• But:– KB still highly incomplete– laborious to build/maintain– difficult to achieve reuse

want a better approach!

Fundamental Problem• Reliance on manual construction of many specific

representations:

– impractical, unmaintainable

– can’t anticipate them all

• But:

– representations contain repeated abstractions• production occurs in photosynthesis, mitosis, growth

• germination includes conversion, production, expansion

• Goal: Capture abstractions in a recomposable way

Bioremediation Amount Amount

Oil Fertilizer

Get Apply BreakDown

Absorb

Microbes Script

Bio-technologist

Soil Rate

environmentcontains

Q+ I- Q-I+

amount

productabsorbed

then

agent

patient agent

scriptpollutant

se

rateagent

then then

product

sesese

patient

remediatoramount

Representation of Bioremediation

agent

Bioremediation Amount Amount

Oil Fertilizer

Get Apply BreakDown

Absorb

Microbes Script

Bio-technologist

Soil Rate

environmentcontains

Q+ I- Q-I+

amount

productabsorbed

then

agent

patient agent

scriptpollutant

se

rateagent

then then

product

sesese

patient

remediator

Conversion Amount Amount

Substance

RateQ+ I- Q-

I+

amountraw-materials

rate

product

Substance

amount

amount

An underlying abstraction...

agent

Bioremediation Amount Amount

Oil Fertilizer

Get Apply BreakDown

Absorb

Microbes Script

Bio-technologist

Soil Rate

environmentcontains

Q+ I- Q-I+

amount

productabsorbed

then

agent

patient agent

scriptpollutant

se

rateagent

then then

product

sesese

remediatoramount

Digest

Substance

BreakDown

Absorb

Agent Script

absorbedagent

script food

se

then

se patient

eater

agent

Another abstraction...

patient

agent

Bioremediation Amount Amount

Oil Fertilizer

Get Apply BreakDown

Absorb

Microbes Script

Bio-technologist

Soil Rate

environmentcontains

Q+ I- Q-I+

amount

productabsorbed

then

agent

patient agent

scriptpollutant

se

rateagent

then then

product

sesese

patient

remediatoramount

Another abstraction...

Treatment

Thing

Get Apply

Thing Script

Person

agent

patient

script patient

agent

then

sese

patient

applied

applied

agent

agent

A Component-Based Approach• Represent component abstractions explicitly• Define concepts as compositions• Construct representations on-demand to answer qns

KB Architecture• Component theories = abstract, reusable models

• Definitions = specifications of compositions

• Inference = construct compositions as needed to answer questions.

Lessons from a Dictionary...

Car: a Vehicle for Passengers

Vehicle: a Means for Transporting something

Transport: to Move from one Place to another

Move: to Go Go: to Move

Specific conceptsdefined as compositions

of abstract concepts

Most abstractconcepts appeal tocore, foundational

theories

1. Component Theories

• A coherent, encapsulated system of concepts & relns

• Contains:– ontology (vocabulary of concepts and relations)– axioms (rules) relating these

• Provides semantics for these concepts in the KB

• Can layer these theories (define one using others)

Example: Distribution NetworkRules (axioms):

• PRODUCERS produce MATERIAL.

• CONSUMERS can consume MATERIAL.

• A network element may be BLOCKED or UNBLOCKED.

• If an element connects with an UNBLOCKED element, then it has an ACCESS to that element.

• A CONSUMER is SUPPLIED if it has ACCESS to a PRODUCER.

• ….

OntologyProducerIntermediaryConsumerMaterialconnectssuppliedstate

Producer Producer

IntermediaryConsumer

Intermediary

Axiom Representation (example)

e1:Elementconnected-

element

Unblockedstatee2:Element

e1:Elementaccess-to

e2:Element

e1,e2:Element connected-element(e1,e2) state(e2,Unblocked) access-to(e1,e2)

(every Element has (access-to ( (allof (the connected-element of Self) where ((the state of It) = Unblocked)))))

“If an element connects to an UNBLOCKED element, then it has ACCESS to that element.”

(iii) Implementation(KM)

(ii) Logic

(i) Semanticnetwork

Other component theories...

• Supply-and-demand

• Containment

• Machines

• Production network

• Two-state object

• Transportation

• ...

2. Definitions and Composition• Definition = specification of a composition

– a Fuel-Cell is a Producer of Electricity

– a Bulb is an Electrical Resistor producing Light

– a Camera is an Image Recording Device

– a Wire is a Conduit of Electricity

• Automated composition:

– Elaboration: component supplies info to answer query

– Classification: recognize concepts in the composition

(Camera has (superclasses (Device)))

(every Camera has (behavior ((a Recording with (input (Image)))))

2. Composition (example)

Composition: Camera = an Image Recording Device

RecordingImage

Devicebehavior

input

Query: Failure modes of a camera?

(Device has (superclasses (Physobj)))

(every Device has (behavior ((a Activity))) (failure-modes ( (the failure-modes of (the participants of the behavior of Self))))))

Activity

Physobj

DeviceFailure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

modebehavior

participants

Physobj

Component Theory: Devices

Physobj Physobj

part.part.

Failure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

modeRecordingImage

Devicebehavior

input

Physobj Physobj

part.part.

Failure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

modeRecordingImage

Devicebehavior

input

Query: Failure modes of a camera?

Sub-query: Participants in its behavior?

Component Theory: Recording(Recording has (superclasses (Activity)))

(every Recording has (input ((a Signal))) (participants ( (a Receptor with (input ((the input of Self))) ...

RecordingSignal

Receptor Memory-Unit

Receiving Writing

Signal

input

agent subevents patient

inputoutput

input

participantparticipant

Physobj Physobj

part.part.

Failure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

modeRecordingImage

Devicebehavior

input

Receptor Memory-Unit

Receiving Writing

Signal

input

agent subevents patient

part.input

outputpart.

Image

Receptor

SignalRecording

Devicebehavior

input

Memory-Unit

Receiving Writing

input

agent subevents patient

part.input

outputpart.

Failure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

mode

Run-Time Classification:Aperture = an Image Receptor

Image

Receptor

SignalRecording

Devicebehavior

input

Memory-Unit

Receiving Writing

input

agent subevents patient

part.input

outputpart.

Failure-Mode

Failure-Mode

failure-mode

failure-mode

failure-mode failure-

mode

inputoutput

failure-mode

Image

Aperture

Image

Blockage

Run-Time Classification:Aperture = an Image Receptor

Aperture - inputs an image - outputs an image - might be blocked - ...

inputoutput

failure-mode

Image

Aperture

Image

Blockage

Run-Time Classification:Aperture = an Image ReceptorFilm= an Image Memory-Unit

Blockage

Recording

Devicebehavior

input

Memory-Unit

Receiving Writing

agent subevents patient

part.inputpart.

Failure-Mode

failure-mode

failure-mode

failure-mode

inputoutput

failure-mode

Image

Aperture

Image

Run-Time Classification:Aperture = an Image ReceptorFilm= an Image Memory-Unit

Film - includes a sheet coated with image-sensitive chemical - might age - ...

Blockage

Recording

Devicebehavior

input

Memory-Unit

Receiving Writing

agent subevents patient

part.inputpart.

Failure-Mode

failure-mode

failure-mode

failure-mode

inputoutput

failure-mode

Image

Aperture

Image

Aging

Filminput

failure-mode

Image

Chemical

sensitive-to

covering

Sheet

parts

Aging

Filminput

failure-mode

Image

Chemical

sensitive-to

covering

Sheet

parts

Blockage

Recording

Devicebehavior

input

Receiving Writing

agent subevents patient

part.inputpart.

failure-mode

failure-mode

failure-mode

inputoutput

failure-mode

Image

Aperture

Image

Aging

Filminput

failure-mode

Chemical

sensitive-to

covering

Sheet

Query: Failure modes of a camera? Blockage, Aging

Sub-query: Participants in its behavior? Aperture, Film

KM> (a Device with (behavior ((a Recording with (input (Image))))))_Device01

KM>

(the failure-modes of _Device01)

(the behavior of _Device01)_Recording01

KM>

Demo...

failure modes of its participants? (from Device) what are the participants? a Receptor and a Memory-Unit. (from Recording)

(classification) [Trace: _Receptor31 classified as a Aperture][Trace: _Memory-Unit32 classified as a Film]

an Aperture and a Film. failure modes of an Aperture and a Film? Blocking, Aging. (from Aperture and Film)

KM> (a Device with (behavior ((a Recording with (input (Image))))))_Device01

KM>

(the failure-modes of _Device01)

(the behavior of _Device01)_Recording01

KM>

Demo...

failure modes of its participants? (from Device) what are the participants? a Receptor and a Memory-Unit. (from Recording)

(classification) [Trace: _Receptor31 classified as a Aperture][Trace: _Memory-Unit32 classified as a Film]

an Aperture and a Film. failure modes of an Aperture and a Film? Blocking, Aging. (from Aperture and Film)

(Blocking Aging)

KM>[Trace: _Writing45 classified as a Exposing](_Receiving45 _Exposing46)

KM>

(the subevents of (the behavior of _Device01))

Other Compositions...• Sound Recording Device (tape recorder)

RecordingSound

Devicebehavior

input

• Sound Producing Device (stereo)

• Vibration Recording Device (seismology)

• Idea Recording Device (palmtop)

• etc.

Compound Concepts are Ubiquitous– Botany:

• photosynthesis• plant material distribution• ...

– Aerospace:• turbine gearbox assembly• case drain fluid• …(43k acronyms!)…

– Sentences also:• “The aircraft overshot the runway.”• “The air-conditioning unit had no power.”• ...

Overall Architecture2. Component

theories(computational clockwork)

ProducerConsumer

Circuit...

Distn.Network

PhysobjMove

Location...

Movement

1. Ontology(conceptual vocabulary)

... ... ...

Thing

Overall Architecture1. Ontology

(conceptual vocabulary)2. Component

theories(computational clockwork)

ProducerConsumer

Circuit...

Distn.Network

PhysobjMove

Location...

Movement

... ... ...

ProducerConsumerCircuit

Thing

Overall Architecture1. Ontology

(conceptual vocabulary)2. Component

theories(computational clockwork)

ProducerConsumer

Circuit...

Distn.Network

PhysobjMove

Location...

Movement

... ... ...

Physobj

Activity

Move

Thing

Overall Architecture1. Ontology

(conceptual vocabulary)3. Definitions and

Descriptions(describe conceptsin terms of others)

2. Componenttheories

(computational clockwork)

ProducerConsumer

Circuit...

Distn.Network

PhysobjMove

Location...

Movement

... ... ...

Thing

Bulb = Light-producing Electrical Consumer

4. Databases ofbasic facts(instances)

Prototype KBS: PHaSE Trainer

Computer + Screen

Laser source

PHaSE KB Architecture

... ... ...

Thing

1. Ontology

2-stateObject

BlckableDAG

DAGProcess

Network

DistrnNetwork

OpticalCircuits

Elec.Circuits

MachineDiscreteevents

2. Componenttheories

PHaSE KB Architecture

BlckableDAG

DAGProcess

Network

DistrnNetwork

OpticalCircuits

Elec.Circuits ... ... ...

Thing

1. Ontology2. Componenttheories

Carousel = a Revolving Case for Storage

3. Definitions andDescriptions

PHaSE KB Architecture

BlckableDAG

DAGProcess

Network

DistrnNetwork

OpticalCircuits

Elec.Circuits ... ... ...

Thing

1. Ontology2. Componenttheories

4. Basic Factsabout PHaSE

PH. ScienceChecklists

PH. CircuitPHaSEphysicalstructure

3. Definitions andDescriptions

Carousel = ...

Example Queries“The parts of the PHaSE control panel?”PHaSE-cover-screw, main power switch, PHaSE MOTOR PWR light, …

“The tool for removing the cover screw?”4.5mm screwdriver

“The possible malfunctions of the PHaSE MOTOR PWR light?” “The motor pwr light has no electricity. (_Absence23) The filament of the motor pwr light is burned out. (_Burned-Out24) The carousel motor is tripped.” (_Tripped25)

“The corrective actions for a tripped motor?” “Toggle the power switch of the carousel motor.”

PHaSE User Interface

PHaSE User Interface

Application:Product Description/eCommerce

... ... ...KB

<product-description> <merchant> <name>Abalt Ltd</name> <location> <city>London</city> <country>UK</country> </location> </merchant> <product> <type>bolt</type> <size>M8</size> <material> <base>titanium</base> <alloy>3Al-2.5V</alloy> </material> </product> <cost> <amount>0.03</amount> <currency>GBP</currency> </cost></product-description>

• “Low-priced fastener?”• “no import restrictions?”• “heat-resistant to 600F?”• “nearby supplier?”• ...

“M8 titanium alloy bolt”

(XML)

Knowledge Requirements for this

... ... ...KB

• “Low-priced fastener?”• “no import restrictions?”• “heat-resistant to 600F?”• “nearby supplier?”• ...

Component Theories

Definitions

Fact databases

Transportation (transport, location, vehicle, …) Commerce (buyer, goods, …)Finance (money, account, exchange-rate, …)Material physics (temperature, density, …)...

Purchase = an exchange of goods for moneyDelivery = the transport of goods from a

seller to a buyer...

GeographyVendorsMaterialsPart-lists

Application: Incident DB Search

950708025099GTHE AIRPLANE OVERSHOT THE RUNWAY. STOPPED 40 FEET FROM END.

961003038219CNUMBER 1 ENGINE FAILED DURING TAKEOFF. RETURNED.

961211044319CA PASSENGER CUSSED OUT THE FLIGHT ATTENDANT. PASSENGERREMOVED.

961203043609CBLEW TIRE DURING LANDING.

FAA Flight Incident Database

Example Search Questions:

• “Which events affected the propulsion? (2)• “Which events might have damaged the undercarriage? (1,4)• “Which events required a mechanic? (1,2,4)• ...

(2)

(1)

(4)

(3)

Application: Incident DB Search

950708025099GTHE AIRPLANE OVERSHOT THE RUNWAY. STOPPED 40 FEET FROM END.

961003038219CNUMBER 1 ENGINE FAILED DURING TAKEOFF. RETURNED.

961211044319CA PASSENGER CUSSED OUT THE FLIGHT ATTENDANT. PASSENGERREMOVED.

961203043609CBLEW TIRE DURING LANDING.

FAA Flight Incident Database

(2)

(1)

(4)

(3)

(a Incident with (aircraft ((a Piper-PA-32))) (destination (OHare-Airport)))) (event ((a Overshooting with (agent ((the aircraft of Self))) (target ((the runway of (the destination of Self))))))

Representation (“Specification”) + Composition Answers

Related Work

• Component-based approaches– Compositional Modeling (CML, Xerox)– Description Logics (composition)– problem-solving methods (KADS)– contexts (Cyc)– s/w engineering (many! Patterns, Comp. Arch)

• Large-scale KBs– Cyc, BKB, TOVE, HPKB– WordNet, Pangloss

Summary• Demand and potential of knowledge processing

• Component-based architecture

– ontology

– core theories

– definitions (specifications of compositions)

– basic fact libraries

• Staged, evaluable development possible

– simple, inferred fact delivery…

– …to a large-scale knowledge resource