Design of knowledge repository

Model for a knowledge repository

To develop a model in which can converge source information from structured
(tabular, XML)
data and unstructured (full text, etc.) the attempt is to provide modelling
constructs for a
propositional context in where statements model facts providing from such
disparate
sources into an unified manner.

Regarding propositional data, statements of a language or facts coming from
tabular
sources can be arranged in the form:

- lover loves loved with Love
- implicant implies implied with implication

And derive from that arrangement that there are two roles (lover, loved), an
action (loves) and a context (love)

Then, regarding Peirce three main semantic primitives:
1. Existence (to be)
2. Relation (to has)
3. Co-reference (state):

Identify previously stated roles as forms of the first primitive:
Existence (to be), "To be a lover".

Identify previously stated contexts as forms of the second semantic
primitive:
Relation (to has), "Lover has love".

And identify the stated actions as an instance of a Verb (to love):
"To love, loving(action), loved(action)" which is to be regarded as the
third primitive,
Co-reference or state.

Then, arrange such relations in a way where they are the 'static' parts,
like classes of
an ontology, and dynamic parts came as instances of them, having:

1. (Roles, Contexts, Verbs) 'static' part
2. (Actors, States, Actions) dynamic part, respectively instances of the
static entities.

Then, just for visualizing what this implies, arrange the statement in a
class-instance
pair manner, like this:

[Verb:to love] [Action: loving] - [Context:love] [State:one love] -
[Role:lover] [Actor:Peter]

[Verb:to love] [Action: loved] - [Context:love] [State:one love] -
[Role:loved] [Actor:Mary]

schema:

[Verb:state] [Action: loving/loved] - [Context:has] [State:love] - [Role:be]
[Actor:lover/loved]

or,

[Verb:to live] [Action: lives] - [Context:life] [State:one life] -
[Role:living] [Actor:Socrates]

[Verb:state] [Action: alive/death] - [Context:has] [State:life] - [Role:be]
[Actor:living/dead]

So, translate in traditional propositional statements:

X to be (is a) lover
X to has (has) love
X state (being) loving

Socrates to be (is a) living
Socrates to has (has) life
Socrates state (being) alive

Note that the verb allows to state that you are on either side of the
relationship or on either
state of the possible values, alive or else of the action of being alive.

Classes (Model over meta model)

This are 'classes' modeled over meta model constructs that should represent
the knowledge
in a propositional manner as it was previously stated, propositional form
suitable for inference or
easily transform into other knowledge representation forms so, for example,
to leverage and take
advantage of their inference mechanisms too. From this object model multiple
forms of statement
or propositional forms can be exported (transformed), and also an
exportation mechanism for
integration with source systems (tabular or structured) or, for example, a
textual output for integration
with input from natural language processing, providing this knowledge
application
as an actor in multiple system or node processes.

Actor(Role, State)
Role(Context, /Actor) : Actor
State(Context, Action) : Role
Context(Verb, /State) : State
Action(Verb, Actor) : Context
Verb(Role, /Action) : Action

Derived (/field name) fields are to be inferred when parsing input data into
the meta model
so building relationships like X is an Actor of Y Role, and State S is a
State of given Context
given an Action (self populating model through inference of relationships).

The main difficulty is to find an 'editable' serialization like HTML for web
pages,
like an XML notation or textual representation.

The full content and related articles, and the model previously presented is
to be implemented
over a meta model which is available in source form from Google Project
page. There is also
the class diagram of the model instances.

http://code.google.com/p/cognescent/

Best,
Sebastian