forgetit: human memory inspired information model
TRANSCRIPT
ForgetIT: human memory inspired Information Model
PERICLES - 27 April 2016Walter Allasia
[email protected] - [email protected]
http://forgetit-project.eu/
Agenda● How human memory works
○ Connecting the memory operation to its computer equivalent
○ Fundamental similarities between human memory and digital preservation: remember=retrieving?
● Main components of the conceptualized Information Model
● Implementing the “preservation value”
How memory worksWhat’s the value of ?
How memory worksWhat’s the value of ?
Answer: 3.14
How memory works
Picture taken from:https://www.trainingindustry.com/wiki/entries/forgetting-curve.aspx
How memory worksWhat’s the value of ?
Answer: 3.1415
How memory works
Picture taken from: http://avidcollegeready.org/college-career-readiness/2011/1/19/why-do-i-have-to-take-notes-the-brain-note-connection.html
How memory works
Picture taken from: http://avidcollegeready.org/college-career-readiness/2011/1/19/why-do-i-have-to-take-notes-the-brain-note-connection.html
Re-Learning!
How memory works
cue:
Cue specific encoding
Math stuff
Geometry stuff
Circle stuff
Spreading Activation Model
Constants
3.1415
This is an example
How memory works
Picture taken from:
http://www.kpetersen.com/rustic-oak-restaurant-tables.htm
What can you eat at a restaurant you have never been to before?
How memory works
Picture taken from:
http://www.kpetersen.com/rustic-oak-restaurant-tables.htm
What can you eat at a restaurant you have never been to before?
Menu on the tables provides the list of available food
How memory works
Picture taken from:
http://www.kpetersen.com/rustic-oak-restaurant-tables.htm
What can you eat at a restaurant you have never been to before?
Menu on the tables provides the list of available food
Pre-existing knowledge is filling the gaps
How memory worksPre-existing knowledge
Knowledge Base
Semantic Memory
Semantic memory example
Hierarchical knowledge base
ImageNet
Biomedical taxonomies
Working memory
Working Memory
ProceduralMemorySemantic
MemoryEpisodicMemory
Personal perspective
Working memory
Picture taken from:http://www.ou.edu/deptcomm/dodjcc/groups/02b1/02b1litreview.htm
Organizational perspective
ForgetIT approach
ForgetIT approach
Personal Contents
Organizational Contents
Preserve or Forget (PoF)
Middleware
Digital Preservation System (DPS)
The PoF Preserve-or-Forget Architecture
The PoF Preserve-or-Forget Architecture
Organizational perspective
Personal perspective
CMIS
Similarities:remember=retrieving?
Processing flow: from Episodic Element to Content
Episodic Element
(date,place,people,context,..)
sensor inputs
Initial Perception
Processing flow: from Episodic Element to Content
take picturesEpisodic
Element
sensor inputs
recordaudio
otherdevices ...
annotations,tags, ...
Initial Perception
Processing flow: from Episodic Element to Content
take pictures
Device Metadata(Date, GPS,
FoV, Light,...)
Episodic Element
sensor inputs
recordaudio
otherdevices ...
annotations,tags, ...
Initial Perception Specific Cue encoding
Extract cue(hook)
Processing flow: from Episodic Element to Content
take pictures
Device Metadata(Date, GPS,
FoV, Light,...)
Episodic Element
sensor inputs
recordaudio
otherdevices ...
annotations,tags, ...
Initial Perception Specific Cue encoding
Extract cue(hook)
Content Processing
Processing flow: updating the knowledge base
BLOB
FeaturesConcepts
Ancillary... Context
BLOB
FeaturesConcepts
Ancillary...
GraphStructures
cueconcept1
concept2
concept3
concept 4
W1,3 W1,4
W2,4
W1,2
Context
Knowledge Base and Currently Activated Associations
Processing flow: updating the knowledge base
BLOB
FeaturesConcepts
Ancillary...
Relational StructuresGraph
Structures
cueconcept1
concept2
concept3
concept 4
W1,3 W1,4
W2,4
W1,2
Context
ClustersOthers...
Knowledge Base and Currently Activated Associations
Processing flow: updating the knowledge base
Graph of concepts in imagescity park
cue/hookc1
Graph of concepts in imagescity park
cue/hooktrees
outdoor
vegetation
building
urban_park
plant
sky
daytime_outdoor
suburban
blue
green
garden
c1
c2
c3
c4
c5
c6
c7
c8
c9
c10
c11
c12
c13
a9,11
Building the knowledge baseScore array
Sample set provided by CERTH
11270 images346 concepts
Building the knowledge baseScore array
Correlation Matrix
Sample set provided by CERTH
11270 images346 concepts
Building the knowledge baseScore array
Correlation Matrix Weighted Adjacency Matrix
Sample set provided by CERTH
11270 images346 concepts
The “Network of concepts”
Statistical analysis of networkThe Semantic Memory represented with this approach can be considered a
“Complex Network”:
- clustering coeff: 0.21- avg path length: 2.8- asymmetric, fat-tailed P(s)
Size of the giant component (Largest Connected Component) in function of the fraction of nodes removed
The Information Model
User Perspective
User Perspective
Middleware (PoF) perspective
Mapping to standards
Computing the Preservation Value
Contents to be preserved
PoF Middleware
Distributed / Cloud
DPS
From texts: extract context, main topics, etc...From images: extract concept, main topics, features, etc…
Personal perspective: PIMO+PoF processorsOrganizational perspective: Typo3+PoF processors
Computing the Preservation Value
Contents to be preserved
PoF Middleware
Distributed / Cloud
DPS
From texts: extract context, main topics, etc...From images: extract concept, main topics, features, etc…
Personal perspective: PIMO+PoF processorsOrganizational perspective: Typo3+PoF processors
If topics or specific features are similar to something already processed, the related contents are suitable to be more important for the user/organization
Computing the Preservation Value
Contents to be preserved
PoF Middleware
Distributed / Cloud
DPS
Time T0
Retention % (human) = preservation value (IS)
Time T1
Thanks for your attentionFurther readings:
http://www.forgetit-project.euAny questions to: