ACTR forays into the semantic web - PowerPoint PPT Presentation

1 / 26
About This Presentation
Title:

ACTR forays into the semantic web

Description:

ACT-R forays into the semantic web. Lael Schooler. Jose Queseda. Hansjoerg. Neth. Martijn ... Hansjoerg. Neth. Progress. Module. Can activation scale to large ... – PowerPoint PPT presentation

Number of Views:32
Avg rating:3.0/5.0
Slides: 27
Provided by: frank395
Category:

less

Transcript and Presenter's Notes

Title: ACTR forays into the semantic web


1
ACT-R forays into the semantic web
  • Lael Schooler

2
Annette ten Teije
Martijn Brakenhoff
Jose Queseda
Frank van Harmelen
Hansjoerg Neth
3
(No Transcript)
4
RDF (Resource Description Framework)
5
Linked Data Semantic Repository
(www.ontotext.com/ldsr
)
6
Deliverable
www.larkc.eu
7
Usecase
  • Urban planning
  • SNP (gene) for predicting disease
  • WHO cancer Monograph production

8
A LarKC Pipeline
Decider
Transformer
Identifier
Selecter
Reasoner
RDF Repository
9
Challenge and Opportunity for ACT-R
  • Bring ACT-R Concepts to Semantic Web
  • Bring the messy world into ACT-R models
  • Can activation scale to millions or billions of
    chunks?
  • Entice ACT-R community to consider Semantic Web
  • Early days

10
Mapping the Semantic Web and ACT-R
Martijn Brakenhoff
  • Similar Representations
  • RDF ACT-R Chunks
  • Similar control problems
  • LarKC Plug-in ACT-R module

11
Linked Data Semantic Repository
(www.ontotext.com/ldsr
)
12
RDF in XML (Extensible Markup Language)
  • ltrdfRDFgt
  • ltfoafPerson rdfabout"MG"gt
  • ltfoafnamegtMonica Gellarlt/foafnamegt
  • ltfoafmbox rdfresource"mailtomonica_at_fri
    ends.com" /gt
  • ltfoafhomepage rdfresource"http//www.fr
    iends.com/monica" /gt
  • ltfoafworkplaceHomepage rdfresource"http//www.
    restaurant.com/chef/" /gt
  • ltfoafknows rdfnodeID"RG" /gt
  • ltfoafknows rdfnodeID"CB" /gt
  • lt/foafPersongt
  • ltfoafPerson rdfabout"RG"gt
  • ltfoafnamegtRachel Greenlt/foafnamegt
  • ltfoafmbox rdfresource"mailtorachel_at_fri
    ends.com" /gt
  • ltfoafhomepage rdfresource"http//www.fr
    iends.com/rachel" /gt
  • ltfoafworkplaceHomepage rdfresource"http//www.
    central-perk.com/waitress/" /gt
  • ltfoafknows rdfnodeID"MG" /gt
  • lt/foafPersongt

ltfoafPerson rdfabout"RG"gt
ltfoafnamegtRachel Greenlt/foafnamegt
ltfoafmbox rdfresource"mailtorachel_at_friends.com
" /gt ltfoafhomepage rdfresource"http//w
ww.friends.com/rachel" /gt ltfoafworkplaceHomepage
rdfresource"http//www.central-perk.com/waitres
s/" /gt ltfoafknows rdfnodeID"MG" /gt
lt/foafPersongt
13
XML ? RDF Triples
ltfoafPerson rdfabout"RG"gt
ltfoafnamegtRachel Greenlt/foafnamegt
ltfoafmbox rdfresource"mailtorachel_at_friends.com
" /gt ltfoafhomepage rdfresource"http//w
ww.friends.com/rachel" /gt ltfoafworkplaceHomepage
rdfresource"http//www.central-perk.com/waitres
s/" /gt ltfoafknows rdfnodeID"MG" /gt
lt/foafPersongt
subject predicate object _node14accvpnbx5
rdftype foafPerson _node14accvpnbx5 foafna
me "Rachel Green" _node14accvpnbx5 foafmbox ltm
ailtorachel_at_friends.comgt _node14accvpnbx5 foafh
omepage lthttp//www.friends.com/rachelgt _node14a
ccvpnbx5 foafworkplaceHomepage lthttp//www.
central-perk.com/waitress/gt _node14accvpnbx5 foaf
knows _node14accvpnbx4
14
RDF Triples
  • subject predicate object
  • _node14accvpnbx4 rdftype foafPerson
  • _node14accvpnbx4 foafname "Monica Gellar"
  • _node14accvpnbx4 foafmbox ltmailtomonica_at_frie
    nds.comgt
  • _node14accvpnbx4 foafhomepage lthttp//www.frie
    nds.com/monicagt
  • _node14accvpnbx4 foafworkplaceHomepage ltht
    tp//www.restaurant.com/chef/gt
  • _node14accvpnbx4 foafknows _node14accvpnbx5
  • _node14accvpnbx4 foafknows _node14accvpnbx6
  • _node14accvpnbx5 rdftype foafPerson
  • _node14accvpnbx5 foafname "Rachel Green"
  • _node14accvpnbx5 foafmbox ltmailtorachel_at_frie
    nds.comgt
  • _node14accvpnbx5 foafhomepage lthttp//www.frie
    nds.com/rachelgt
  • _node14accvpnbx5 foafworkplaceHomepage ltht
    tp//www.central-perk.com/waitress/gt
  • _node14accvpnbx5 foafknows _node14accvpnbx4
  • _node14accvpnbx6 rdftype foafPerson
  • _node14accvpnbx6 foafname "Chandler Bing"

15
RDF Graph
...
...
node14accvpnbx5
Rachel Green
foafname
foafknows
...
foafknows
...
node14accvpnbx6
Chandler Bing
foafname
...
...
16
ACT-R Chunks
  • (_node14accvpnbx4
  • ISA foafPerson
  • foafname Monica_Gellar
  • foafmbox mailtomonica_at_friends.com
  • foafhomepage http//www.friends.com/monica
  • foafworkplaceHomepage http//www.restaurant.com/
    chef/)
  • (_node14accvpnbx5
  • ISA foafPerson
  • foafname Rachel_Green
  • foafmbox mailtorachel_at_friends.com
  • foafhomepage http//www.friends.com/rachel
  • foafworkplaceHomepage http//www.central-perk.co
    m/waitress/)
  • (_node14accvpnbx6
  • ISA foafPerson
  • foafname Chandler_Bing
  • foafmbox mailtochandler_at_s60.com
  • foafhomepage http//www.friends.com/chandler
  • foafworkplaceHomepage http//www.corporation.com
    /statistical-analist/)

17
Chunks are Subgraphs
(_node14accvpnbx4 ISA foafPerson foafname
Monica_Gellar foafmbox mailtomonica_at_friends.co
m foafhomepage http//www.friends.com/monica
foafworkplaceHomepage http//www.restaurant.com
/chef/)
What unit of knowledge should take
activation? Triple? Chunk?
18
ACT-R Retrieval Requests in SPARQL (SPARQL
Protocol and RDF Query Language)
  • (p find-person
  • goalgt
  • ISA find
  • person person
  • gt
  • SPARQLgt
  • ISA foafPerson
  • name person)

SELECT WHERE ?x rdftype foafPerson
?x foafname ?name ?x foafname
Monica_Gellar ?x foafmbox ?mail ?x
foafhomepage ?page ?x foafworkplaceHomepage
?workpage
19
SPARQL Module(was/is working)
20
ACT-R to Orchestrate Pipelines(in progress)
Goal Module
Imaginal Module
Procedural Module
Reasoner Module
Declarative Module
Reasoner
Progress Module
Pipeline Module
Decider
getProgressReports
Transformer
Identifier
Selecter
Reasoner
postProgressReports
RDF Repository
21
Foraging Stopping Rules
Progress Module
Hansjoerg Neth
22
Can activation scale to large numbers of chunks?
23
SELECTION by Centrality
Jose Queseda
15 of IMDb 80 of interest
Cumulative Proportion of Votes
Number of movies selected
24
Challenge and Opportunity for ACT-R
  • Similar Representations
  • RDF ACT-R Chunks
  • Similar control problems
  • LarKC Plug-in ACT-R module
  • Bring ACT-R Concepts to Semantic Web
  • Bring the messy world into ACT-R models
  • Can activation scale to millions or billions of
    chunks?
  • Entice ACT-R community to consider Semantic Web
  • Early days

25
ACT-R forays into the semantic web
  • Lael Schooler

26
ACT-R Workshop Schedule Opening ACT-R from
CMUs Perspective 900 - 945 Overview of
ACT-R -- John R. Anderson 945 1030 Details
of ACT-R 6.0 -- Dan Bothell Break 1030
1100 Presentations 1 Architecture 1100 1130
Functional constraints on architectural
mechanisms -- Christian Lebiere 1130 1200
Retrieval by Accumulating Evidence in ACT-R --
Leendert van Maanen 1200 1230 A mechanism
for decisions in the absence of prior reward --
Vladislav D. Veksler Lunch 1230
130 Presentations 2 Extensions 130 200
ACT-R forays into the semantic web -- Lael J.
Schooler 200 230 Making Models Tired A
Module for Fatigue -- Glenn F. Gunzelmann 230
300 Acting outside the box Truly embodied
ACT-R -- Anthony Harrison 300 - 330
Interfacing ACT-R with different types of
environments and with different techniques
Issues and Suggestions.-- Michael J.
Schoelles Break 330 400 Panel 400 530
Future of ACT-R from a non-CMU Perspective
Danilo Fum, Kevin A. Gluck, Wayne D. Gray, Niels
A. Taatgen, J. Gregory Trafton, Richard M. Young
Write a Comment
User Comments (0)
About PowerShow.com