Cse@buffalo S.C. Shapiro An Introduction to SNePS 3 Stuart C. Shapiro Department of Computer Science...
-
date post
22-Dec-2015 -
Category
Documents
-
view
216 -
download
0
Transcript of Cse@buffalo S.C. Shapiro An Introduction to SNePS 3 Stuart C. Shapiro Department of Computer Science...
cse@buff
alo
S.C. Shapiro
An Introduction to SNePS 3
Stuart C. Shapiro
Department of Computer Science and Engineering
and Center for Cognitive Science
State University of New York at Buffalo
S.C. Shapiro
cse@buff
alo
Outline
• Setting
• Basic SNePS Principles
• Examples
• 4 Kinds of Inference
• Summary
S.C. Shapiro
cse@buff
alo
Parentage of SNePS 3
• SNePS 2.5
• ANALOG– Structured (Conceptually Complete) Variables
• Currently being implemented– in CLOS and/or Java.
S.C. Shapiro
cse@buff
alo
SNePS KRR Style
• Network-based
• Logic-based
• Intended as the LOT of a NL-competent cognitive agent.
S.C. Shapiro
cse@buff
alo
Outline
• Setting
• Basic SNePS Principles
• Examples
• 4 Kinds of Inference
• Summary
S.C. Shapiro
cse@buff
alo
Basic SNePS PrinciplesA Summary of Syntax and Semantics
• Propositional Semantic Network
• Term Logic
• Intensional Representation
• Uniqueness Principle
• Paraconsistent Logic.
S.C. Shapiro
cse@buff
alo
Propositional Semantic Network
• The only well-formed SNePS expressions are nodes.– Arcs do not have semantics
• Do not have assertional import
S.C. Shapiro
cse@buff
alo
Term Logic
• Every well-formed SNePS expression is a term.– Even propositions are denoted by terms.– Propositions can be arguments without leaving first-
order logic.
S.C. Shapiro
cse@buff
alo
Intensional Representation
• SNePS terms represent (denote) intensional (mental) entities.– Cognitively distinct entities denoted by distinct terms
• Even if co-extensional
– Every term denotes a mental entity.• No term for purely technical reasons
S.C. Shapiro
cse@buff
alo
Uniqueness Principle
• No two SNePS terms denote the same entity.– Syntactically distinct terms are semantically distinct.– Full structure sharing.
S.C. Shapiro
cse@buff
alo
Paraconsistent Logic
• A contradiction does not imply anything whatsoever.– A contradiction in one subdomain does not
corrupt another.
S.C. Shapiro
cse@buff
alo
Outline
• Setting
• Basic SNePS Principles
• Examples
• 4 Kinds of Inference
• Summary
S.C. Shapiro
cse@buff
alo
Example: Term Logic& Conceptual Relations
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Example SNePS Ontology
S.C. Shapiro
cse@buff
alo
Cassie talks to Stu
S.C. Shapiro
cse@buff
alo
Outline
• Setting
• Basic SNePS Principles
• Examples
• 4 Kinds of Inference
• Summary
S.C. Shapiro
cse@buff
alo
Wire-Based Inference
S.C. Shapiro
cse@buff
alo
Wire-Based Inference
S.C. Shapiro
cse@buff
alo
Path-Based Inference
S.C. Shapiro
cse@buff
alo
Path-Based Inference
member
class
class
S.C. Shapiro
cse@buff
alo
Path-Based Inference
S.C. Shapiro
cse@buff
alo
Node-Based Inference
If B1 is a talking robot,then B1 is intelligent.
S.C. Shapiro
cse@buff
alo
Node-Based Inference
S.C. Shapiro
cse@buff
alo
Node-Based Inference
S.C. Shapiro
cse@buff
alo
Node-Based Inference
S.C. Shapiro
cse@buff
alo
SNePS 2.5 Generic Version
S.C. Shapiro
cse@buff
alo
Subsumption Inference
S.C. Shapiro
cse@buff
alo
Outline
• Setting
• Basic SNePS Principles
• Examples
• 4 Kinds of Inference
• Summary
S.C. Shapiro
cse@buff
alo
Summary
• SNePS: a Logic- and Network-Based KRR• with its own Syntax, Semantics, Proof Theory• SNePS 3 has 4 kinds of inference
– Wire-based– Path-based– Node-based– Subsumption
• SNePS 3 is currently being implemented.
S.C. Shapiro
cse@buff
alo
SNeRG Home Page
http://www.cse.buffalo.edu/sneps/
S.C. Shapiro
cse@buff
alo
Wire-Based Inference
• Assume
(define-relation
:name “member”
:type entity
:adjust reduce
:limit 1)
S.C. Shapiro
cse@buff
alo
Path-Based Inference
• Assume
(define-path
class
(compose class
(kstar (compose
subclass- !
superclass)))
S.C. Shapiro
cse@buff
alo
Node-Based Inference
E.g. Using and-entailment
{P1, …, Pn} &=> {Q1, …, Qm}