One
FUNCTION: The core architecture has a distributed, virtual, database
with informational transitions and ontology (knowledge of processes).
Client information is maintained locally and communicated to other
locations on a just in time basis. However, clinical medical ontology
is shared globally. The distributed core architecture allows medical
staff access to local doctor / patient records, while still being
responsive to the system need for reporting (to medical insurance
claims, to regional governmental authorities and to the medical
research community) and analysis.
ARCHITECTURE: The architecture is based on a model of information flow
between thousands of locations using state gesture pairing. (a
knowledge operating system working on state gesture pairing is
available). A common semiotic language (sign system) is responsive to
substructure (statistics) and ultra structure (category theory) is
shared throughout the entire HealthCare Infrastructure for a Small
Nation.
TECHNOLOGY: Knowledge engineering and information retrieval
technologies are to be used at each of seven interfaces. The interfaces
are between (1) the doctor or staff to a full text database via voice
and typing input, (2) text to context conversion suing knowledge
engineering techniques, (3) translation between context representation
and a local form based reasoning system based on finite-state
constructs, 4) translation between the local system and a global
reasoning system based on finite-state constructs, (5) archived in a
form based representational system, (6) movement between warehouse and
stores, and (7) movement between stores and the community.
Two
FUNCTION: The system learns user categories and reporting decisions
using a linguistic interpretation of retrieval requests and a separate
SQL based data retrieval paradigm.
ARCHITECTURE: The architecture for learning and reinforcement depends
on an open loop between the community and the global finite-state
construct. This loop enables user requests for information to make
small alterations in a categorization policy that is used to
re-organize some part of the contents of a large virtual text+data
warehouse for display as a local information store.
Three
FUNCTION 1: It is necessary to learn (extract/represent) detailed
knowledge that is required to make judgments about the viability of
goals and intended actions.
FUNCTION 2: It is necessary to learn (extract/represent) detailed
knowledge required to advise clinic workers regarding the standard
practice for any medical situation.
ARCHITECTURE 1: The architecture for nationally centric investigation
is modular and taps into the finite-state constructs so that knowledge
management, artificial intelligence and data mining tools can be
brought to bear in the acquisition phase of an medical investigation.
ARCHITECTURE 2: The architecture for clinic centric investigations
works to bring information to the clinic desktop regarding
relationships that are discovered in the local finite-state construct
or in a global construct consisting of standard medical clinical
ontology.
Four
FUNCTION: The means through which patient information is gathered is via
(1) polling and survey instruments,
(2) non-structured interview, and
(3) examination.
ARCHITECTURE: The polling and survey instruments are immediately
transformed into local finite-state constructs.
Information from the interview is placed into standard forms that are
automatically transformabled into knowledge base constructs as
ontology. Knowledge validation is handled by thesaurus (via knowledge
engineering), linguistic analysis and category theory; plus dialog
between the system and the user.
Information gathered during a doctor's examination of a patient will
undergo knowledge validation.
Five
FUNCTION: Voice to text translation undergoes a checking procedure. The
standard medical clinical ontology will be used in a dialog fashion to
confirm the translation's meaning.
HOW: Dialog is supported using gestures in response to state
approximations of the user's meaning. This state-gesture mechanism is
the core to a Knowledge Operating System (KOS).
Six
ARCHITECTURE: Additional information validation can be done after the
doctor has left the patient and is able to use some interactive tools
to make additional comments, review and analyze the data as stored in a
local construct (ontology).
TECHNOLOGY: Catgorical technology is allowed to organize data into
what-if scenarios.
Seven
FUNCTION: The difference between an ad hoc method and a formal method
has to do with how communication can occur with very specific forms.
The ad hoc method recognizes that pre-determined forms are often not
able to hold the information that is necessary to encodeknowledge from
an interview or examination of a patient. The formal method recognizes
that context non-specific information is formed via abstraction and
generalization.
TECHNOLOGY: Some category technology is required to place text into
context, or to generate the regular forms of context that allows one to
define finite-state constructs and provide for validation of knowledge
base constructs and case reasoning.
Eight
FUNCTION: The information that is placed into the central warehouse can
be an abstraction of actual data, as well as full indices that point to
specific meaningful interpretations of points in the constructs.
ARCHITECTURE: The transition between the global constructs and the
warehouse is via abstraction and concept representation.
FUNCTION: The finite-state constructs have two equivalent forms. The
first is a normalized database (XML based) whose structure is
established by architectural design and data modeling. The second is an
standard medical clinical ontology.
The local / global interface works between the data model and
ontolgoies of the local constructs and the data model and ontologies of
the global constructs. There is no non-computational aspect of this
interface and thus the communication can be optimized for web
transmission.
ARCHITECTURE: The interface between the local and global constructs is
via computations on indices, and thus most often does not involve the
transmission of specific data. This architectural feature has
implication for data compression and security, since the data that is
shared is not about the details of client records, but rather about
pointers to this detail and abstract categorical information that is
statistical in nature.
Nine
FUNCTION: The function of the warehouse-to-store transformation is to
provide only the type of information that each particular user has
expressed an interest in. Information coherence and consistancy is also
evaluated locally using theorem proving techniques.
ARCHITECTURE: The architecture of the transformations between
warehouse, store and community is based on a projection from a large
amount of information to a smaller amount of information. SQL is used
exclusively since the projection is a simple view of the warehouse.
TECHNOLOGY: The store contents can be dynamically configured for each
individual using a database backend that generates HTML for web
browsers.
Ten
Blank
Eleven
FUNCTION: The transformation of information in the local store to
specific users can be made
(1) adaptive (such as with the voting procedure),
(2) seek and fill (such as smart agents frame filling), and
(3) descriptive (such as Autonomy Inc. Dynamic Reasoning Engines).
ARCHITECTURE: The local store must get it's information and information
structures from the warehouse.
However a routing or frame filling retrieval technology may be used to
"purchase" specific information from the store.
Twelve
FUNCTION: Learning and reinforcement technology is used in a very
narrow and specific fashion to make changes in the local constructs.
This is local and global learning, where the changes that are being
made are for the purpose of creating a better categorization and
representation of basic data structures.
ARCHITECTURE: Learning introduces significant modifications to data
structures, particularly in the themespace and in indices.
TECHNOLOGY: The technology here uses a simple weight profile to
retrieve data. The weight profile can be made adaptive thougth the
introduction of neural networks, the voting procedure, and full text
retrieval methods based on the industry standards.