semantic-based knowledge representation 2...signifiers – signs or symbols if you’re a...

SEMANTIC-BASED KNOWLEDGE REPRESENTATION

James C. Trawick

NASA IV&V Program

Human Exploration and Operations – Exploration Systems Directorate

Kennedy Space Center, FL, USA

011.321.861.8396 – [email protected]

IV&V Program

Introduction – What is Semantics? How does Semantics relate to knowledge

representation? How would Semantics be used in the context

of IV&V?Tracking and mining of findings/issues/risksTracking and mining of IV&V evidenceAssurance statement development

What are other uses of Semantic-BasedKnowledge Representation?

What’s next for IV&V?

2

IV&V Program

(as the lawyers say) It DependsThe literal meaning of the Greek word σημαντικός –

signifiers – signs or symbols

If you’re a philosopher

The study of meanings – context

If you’re a Linguist

Linguistic semantics is the study of meaning that is used forunderstanding human expression through language

If you’re a computer scientist

The processes a computer follows when executing a program –the model of computation in a particular programminglanguage

If you’re an IV&V analyst (a little of all of these)

Finding the meaning of evidence in its context

3

IV&V Program






4

IV&V Program

Used when knowledge best understood as a set ofconcepts that are related to one another

Context is everything

NOT a new concept

Extended Ontological/Taxonomic structure makes itderivable, searchable and mineable

Stored relationships provide the data needed toautomatically generate visual network diagrams andto later mine for significances of subtle relationships

5

IV&V Program

6

Vertebra Cat Fur

hasis a has

is an is a

has

MammalAnimal Bear

Whale

Fish Waterlives in

is a

is an

lives in

A Simplistic Semantic Network Model

IV&V Program






7

IV&V Program

Reported anomalies (findings/issues/risks) areinherently contextual (e.g., a software errorwhich would cause a hypergolic propellantvalve to open when it is not supposed to):Catastrophic when the vehicle is being fueled (loss of

life and mission assets)Merely an annoyance when the vehicle is unfueled and

being processed

A semantic-based knowledge system cansuggest appropriate scoring of anomalousfindings based on context and a pre-determined “bar”

8

IV&V Program






9

IV&V Program

A semantic-based knowledge system candescribe evidence element relationships andderive their significance

A semantic-based knowledge system cansuggest additional evidence that must beacquired in order to complete an analysis

A knowledge-based system always generatesa decision path which can be evaluated andlearned from

10

IV&V Program






11

IV&V Program

A semantic-based knowledge system cansuggest assurances that can be made forcaptured assurance element relationships

A semantic-based knowledge system cansuggest caveats that must be attached toassurances that are desired to be made,based on missing assurance elementrelationships

A knowledge-based system always generatesa decision path which can be evaluated andlearned from

12

IV&V Program

13

Assurance Case 3

is assumedby

Assurance Case 1

Test ResultsAnalysis

Assumption

Assurance Case 2

A Simplistic Semantic Network Model of an IV&V Assurance Case

Assurance Case 4

is evidencefor

is partialevidence for

is evidencefor

is partialevidence for

IV&V Program






14

IV&V Program

IBM’s Watson computer systemShowcased on the Jeopardy network television

program

Used as an open-domain question answeringsystem using natural language input

An example of a semantic-based knowledgerepresentation and retrieval system

Competitors – two other semantic-basedknowledge representation and retrieval systems:two well-read humans

15

IV&V Program

16

PrimarySearch

CandidateAnswer

Generation

SupportingEvidenceRetrieval

DeepEvidenceScoring

Question

HypothesisGeneration

Hypothesis andEvidence Scoring

SoftFiltering

HypothesisGeneration

Hypothesis andEvidence Scoring

SoftFiltering

Answer/Confidence

Final Mergingand Scoring

Synthesis

TrainedModels

QuestionAnalysis

QueryDecomposition

Simplified Context Diagram

–

Watson Open

-

Domain Question Answering Engine

EvidenceSources

AnswerSources

IV&V Program






17

IV&V Program

Conclusions

Several semantic-based knowledge systems showpromise for documenting and mining aspects of theIndependent Verification and Validation of software

Semantic-based knowledge systems show promisefor documenting and mining aspects of other areasoutside of the Independent Verification andValidation of software

18

IV&V Program

Suggestions for the future of IV&V

Development of a self-populating database tool forentering entities and relationships in multiplefunctionalities/domains (e.g., requirements quality in aspacecraft guidance, navigation and control [GN&C]domain)

Development of a self-populating database tool forentering ISO/IEC 15026-2 assurance elements and theirrelationships

Development of a customizable common notation forISO/IEC 15026-2 (for display)

Development of display, report generation, analysis andmining tools to make the databases useful

Using the tools to analyze patterns and trends to improvethe IV&V process

19

semantic-based knowledge representation 2...signifiers – signs or symbols if you’re a...

Documents