Designing vectorbased ontologies: Can technology empower open interpretation of cultural heritage

1
Designing vector-based ontologies Can technology
empower open interpretation of cultural heritage?

• Lily Diaz
• Mauri Kaipainen
• Media Laboratory
• Unversity of Art and Design Hesinki

2
Context

• Initial stage
• Collaboration between design and cognitive
  science
• Collaboration of two research groups
• Initial exploration and definition of problem area

3
Conventional and native information handling

• Everyday information technology Things come with
  category tags on them.
• Native information-handling, neuro-inspired
  information handling
• There is more to mental maps than metaphor
• The brain projects rich informations from the
  body onto the cortex

4
Evidence for native information handling

• Brain imaging techniques EEG, MEG, MRi
• Distributed, focally local activity patterns
  Tonotopies, somatotopies
• Artificial neural networks simulate adaptive
  neural mechanisms. Existence proof.
• Statistical methods multidimensional scaling
• Everyday experience Class memberships are fuzzy
  (How tall is a tall person?), class judgments
  vary depending on perspective.

5
Interpretation and projection

• Class membership is perspective-dependent
• All projections involve a transformation of data.
• Refers to the process of structuring knowledge
  based on acquired knowledge.
• Interpretation as projection

6
Projected constructs in archaeology

• Colin Renfrew (1995 10) argues that the idea of
  the past is related to an individuals
  acquisition of knowledge about the world, and her
  ability to formulate projected constructs or
  models about its nature.

7
Brain-inspired design of information systems

• Premise Brain and mind are fundamentally same
  (naturalist / physicalist standpoint)
• Hypothesis Better cognitive ergonomy can be
  achieved using brain-like information-handling.
• Implication Data is expressed in a form as
  elementary as possible to allow bottom-up,
  data-based emergence of similarity clusters
  (self-organization)

8
Art for the mind and the brain

• Solso (1994)
• Image construction operates according to the
  principles of human information processing.
• The artist finds expressions, patterns, that
  stimulate the human neural system.
• The artist creates art for the mind.
• Art is an integration of perceptual skill and
  higher order cognition that engages past
  knowledge and interpretation.

9
Ontological spaces

• Rethinking ontologies Elementary descriptive
  properties define an ontological space of an
  artefact
• The artefacts position in the ontological space
  can be projected onto an interpretative layer in
  a number of ways.

10
Numeric vector-based ontologies, numeric

• Artifact represented by a vector, with each of
  its N components,
• corresponding to the degree of presence,
  relevance or probability of a specified
  elementary property,
• expressible in terms of a value between 0 and 1
• e.g. youngness .75, boldnes 0,
  lot-of-rain-in-winter .1
• Drawback all N components have tobe fixed

11
Vector-based ontologies, Text-based method

• Honkela et al. 1996
• Artifact represented by the sum of vectors, each
  of which stands for an elementary property
• Each property vector randomly assigned
• With a frequent property, the correspondent
  vector is multiplied proportionally to frequency
  of occurrence
• New properties can be introduced if necessary

12
Ethical consequence Open Interpretation approach

• It need not be the task of the information
  designer to chew the world for the user. The
  designer should respect the freedom of the user
  to interpret the environment in her own way.
• The designer should provide support for native
  information handling.

13
Design in the Age of Information

• Report to the National Science Foundation,
  Krippendorf et. al. (1997)
• Current trends seek to elaborate a discipline of
  design into a second-order science that attempts
  to find out how different groups or cultures
  possess particular ways of conceptualizing their
  practices and the world.

14
Cultural heritage

• Collaborations between the arts, humanities and
  sciences represent a unique opportunity, not only
  to form new partnerships, but also to develop new
  products and new design approaches.
• Items of culture heritage are complex objects
  synthesizing diverse conditions of knowledge
• Conceptual framework
• Material conditions

15
Objects of culture

• Whether immaterial or physical, all objects are
  manufactured.
• Wartofsky 1979, Cole 1986
• Primary artefacts
• Secondary artefacts
• Tertiary artefacts

16
Case Relaciones Geográficas

• rich record of living conditions in early
  colonial America
• systematical
• a fixed number of questions
• historically simultaneous, mutually comparable
  data
• good geographical coverage
• 208 locations, enough data not to offer too
  obvious interpretations
• primary sources

17
Primary analysis of a sample of questions

• 15 first questions explored
• How can data be translated into elementary
  properties (dimensions)?

18
Conclusion of the primary analysis

• Descriptions of climate (3), landscape (4),
  indigeneous population (6), city location (10),
  and anthropological descriptions (14) can be
  broken into elementary properties general enough.
  

19
Conclusion of the primary analysis (continued)

• Distances between cities (7. 8. 11. 12) translate
  into matrices that could be used to reconstruct
  the experienced geography/mental map
• Questions relating to unique names or narratives
  (1, 9, 11, 13), or histories (2) do not define
  shared ontological space gt have to be described
  by other means.

20
Conclusion of the primary analysis (continued)

• Ontological space description is only partly
  applicable
• Summa summarum Essential aspects of the data
  provide a basis for an ontological space
  description and will allow for a rich freedom for
  open interpretation,
• given a versatile enough toolset for accessing
  the data

21
Tools for Interfacing with ontological spaces

• Background work (Kerminen et al. 2000, Kaipainen
  et al. 2001)
• Visualization
• Viewing options
• Interpretative lens
• Naming tools
• Linking tools

22
Discussion

• Already existing large base of culture heritage
  objects in electronic digital format.
• This will be a continuing trend.
• Need is for new methods that allow us to
  restructure the types, modes and roles of
  information ecology.

23
Discussion (continued)

• It does not have to be the researchers or
  knowledge designers or engineers task to define
  the categories.
• Can the vector-based method of defining
  ontological spaces of human artifacts provide an
  alternative to conventional ontologies?
• Defending the right of the information user for
  her open interpretation.

24
Discussion (continued)

• Neurophysiology may indicate ways in which
  information technology may be drawn closer to the
  native cognition,
• relying on the holistic capabilities of the
  cognition to structure things in a spatial manner
• Initial sample of different knowledge building
  approaches seem to support this proposition.
• Archaeoloy
• Art

25
Discussion (continued)

• Relaciones Geográficas
• Richness of data is broken into its elementary
  properties.
• Ontological space allow for a number of different
  projections
• Open interpretation approach is based on
  subject-given weights of different dimensions on
  the space.

26
Open questions

• Virtues and problems of open approach are still
  to be explored.
• Interface solutions are a challenging task that
  we have begun to work on it.
• Can the user be constituted as an open ended
  category as well?

27
Open questions (continued)

• Can such system be designed and implemented with
  low overhead costs?
• Can user ergonomy be enhanced by neuro-inspired
  computation?
• Is there a need for a soft technology that leaves
  decisions up to the user, or is it that we use
  computers exactly to get away from out own native
  softness?