Information Technology Reference
In-Depth Information
5.11.4 Revision with Frame
Frame, as a knowledge representation method, is suggested by Marvin
Minsky for recognition, natural language dialog, and other complex actions in
1975 (Minsky, 1975). A frame is a data-structure for representing a
stereotyped entity, such as a situation, a concept, or an event. As for a new
situation, we can get an instance of the entity by filling the values to the
structure. The frame structure also offers a way to prediction for a particular
entity under a concrete context where we can find the needed information.
Slots are the places in frame structures where predication information stored.
A frame comprises a number of slots, and each slot has its own name and the
values that go into it. The values in the slot describe the attributes of the
entity the frame represents. Usually, the slots are complex structures that have
facets describing the properties of them, and every facet describes the
characteristic of the slot from one viewpoint. Each facet takes one or more
value, which is described by a concept or some concepts. Frame theory is a
important knowledge representation technology and gains extensive
application in that it organize knowledge in a human-like and concise manner.
But the existing frame systems still lack the way to deal with conflicts in
prediction and uncertain knowledge.
According to the prediction and study on marine fishery resources, we
defined a frame system KBIF with a power to express uncertain knowledge
with different importance. The main characteristics of KBIF includes: Weight
- the influence degree of every slot on the frame, trust factor - the degree of
every slot can be believed, and preference factor - the inclination degree in
inheritance from parent frames.
The fish migration is influenced by a lot of factors and is too complicated
to be described by traditional mathematics method and model. The prediction
algorithm in CBR can only make predication based on some factors
accumulated through a large number of historical data, while the expert
knowledge of the fishing grounds is inaccuracy and incomplete. The key point
in prediction accuracy is how to revise the prediction results about the central
fishing grounds effectively. The knowledge representation in
adopts
three kinds of knowledge processing models: frame model, blackboard
structure, fuzzy reasoning, and accordingly expert revision systems are divide
into three parts: atom set, rule set, and conclusion set. The application of three
KBIF
Search WWH ::




Custom Search