The PRISM pattern recognition software is able to recognize distorted
images which have been rotated or scaled in size. It can recognize
different image elements randomly scattered throughout the picture.
This is because each image is represented by individual topological
picture elements called isoterms which are composed of alphanumeric
These isotermal sequences can be placed in a dictionary for lookup
identification. In the case of textual identification, sets of
words are the patterns which are put in the dictionary. The elements
of any sequence representing pictures, sounds or text can have logical
relationships. The profiles for each pattern can be constructed
with logic control records as explained below.
The profiles compiler, PC, condenses sets of pattern descriptors called
profiles into a sets of associative lists which are used to perform pattern
recognition. A profile is composed of a logical sequence of character
strings describing the pattern data. These string sets are be group into
composite elements which may be transformed with boolean operators.
The result is a profile which can represent a complex pattern. The
PC condenses many profiles together in an efficient manner.
The Attribute Filter, AF, uses the dictionary produced by the PC module
to evaluate the text data and assign attribute codes to the data.
Single words or sequences of words are the elements which PRISM uses
to classify data. These words are nodes in a semantic network. The mechanism
which facilitates this semantic description is the profile.
logic: A + B.
A: item1, item2 item3, item4;
Single words or a sequence of words are the attributes which PRISM uses
to classify data. Data descriptors in a semantic network are linked to
other elements in the net thereby forming relationships. These sets or
sequences of words are used as composite entities for logic processing.
They are merely called operands. These operands are used to build the
logical relationships in the semantic network. The text description
can be condensed in a profile or topic. The profile contains
the attribute of the text data you want to filter. Furthermore, a profile's
logical disposition is defined in relationship to all the other profiles.
A profile is node in the semantic network. The profile is formally defined
by the following items.
(1) Node name which is an element define
globally across the entire network.
(2) Logic operands which are nodal elememts
defined system wide.
(3) Description of each operand consisting
of keywords and phrases.
(4) Association consists of words describing
the semantic connections.
The profile or nodal element can represent an image, idea or concept
defined by logic functions or relationships, attribute descriptions,
and associating connections.
A specific example is of the form:
logic: A but not B.
definition: A: lion, zebra, hyena;
B: panda, kangarro.