Abduction
A form of reasoning by which, given a fact or state, we can hypothesize some prior fact or state which might have given rise to it. For example, if we know that the streets are always wet after it has been raining, and that the streets are wet now, we can guess that it has recently been raining. The streets could be wet, of course, for some other reason entirely, so abduction is not always a reliable form of reasoning. See also Logic, Rules of Inference.
Agenda
A list of possible tasks for a system to perform, ordered
according to usefulness. An agenda may be used, for example, in a
search program to determine the next path to investigate.
Algorithm
A fixed set of operations on numbers or symbols - of a
sort that can be automated on a digital computer. See also
Heuristic.
Analogue computer
See Digital/Analogue Computer.
Analogical
In an analogical representation there is a direct
correspondence between the important properties of the representation and
the properties of the domain -- for example, by using a 2-dimensional
array to represent a 2-dimensional picture.
Arc/Link/Pointer
Nodes in a semantic network are
connected by labelled directed arcs (also sometimes called links or
pointers), which indicate the nature of the connection between adjacent
nodes. In isa-hierarchies, the most basic links are taxonomic (isa)
and partonomic (ispart) relations; others name features or attributes; their
values are given by the node pointed to by the link.
Architecture
(As applied to computers), the structured
organization of elements in a computational mechanism -- for example, the
difference between sequential processing and
parallel distributed processing machines is said to be a difference in
architecture. See also Computation,
(non-) von Neumann Architecture.
Argument/Result
A procedure may take input values and
produce output values. The former are called the procedure's
arguments; the latter its results. Thus the POP-11
procedure call length([1 2 3]) has one argument (the list
[1 2 3]) and produces one result (the number 3). In logic,
an argument represents individual entities, such as `John', or `apple'. A
predicate is applied to one or more arguments to produce a
formula. See also Axiom/Theorem.
Array
A data object consisting of an ordered series of elements
that can be accessed by number index.
ASCII Code
The widely used American Standard Code for Information
Interchange (ASCII Code) assigns a number to each character that can be
typed on a computer keyboard. The code aids programming and
communication between computers.
Assignment
The operation in computer programming of giving a
value to a variable.
Attribute/Feature/Property
An attribute (or feature, or property) is
a distinguishing aspect, quality, or characteristic of an object. An
attribute of a canary, for example, is yellowness.
Axiom/Theorem
Most logical systems contain fundamental propositions
or axioms, which cannot themselves be justified from within the system
itself. Any conclusions derived from these axioms by the system's
rules of inference are called theorems. See also Logic,
Predicate Calculus/Predicate Logic, Propositional Calculus.
Backtracking
Returning to a previous state in a computation in order
to try an alternative strategy. In terms of a search tree,
backtracking occurs when the end of a branch is reached without a solution
being found, and a different branch is chosen.
Backup Memory
See Main Memory/Backup Memory.
Behaviourism/Behavioural Psychology
An influential twentieth-century
school of thought in psychology and in the philosophy of mind that claimed
that only the outwardly observable aspects of human behaviour produced
scientific data for the study of the mind. Behaviourists thus rejected
introspectionism.
Binary Digit (Bit)
Most modern computers use the binary
counting system -- employing only 1's and 0's -- to represent any number
or other character. Binary numbers can be easily stored in electronic form.
See also Central Processor.
Boolean
A data object which represents whether some condition
is true or not. There are only two such objects: <true> and
<false>.
Branch and Bound
A search strategy in which the `lowest cost'
node in the agenda is always considered first. This strategy is
guaranteed to find the lowest cost solution if more than one solution to a
problem exists.
Branching
See Conditional.
Breadth-First
A search strategy in which each level of the
search tree is considered in turn. All the nodes at any given
level of the search tree are considered before any nodes at a deeper level.
Built-In/User-Defined
Most programming languages have a variety of
terms which are primitive to the language; these are the built-in
terms. Many languages, such as POP-11, allow you to define your own
terms or procedures. These are known as user-defined terms.
Calling a Procedure
Instructing a computer to interpret
the series of instruction defined in a procedure. See also
Argument/Result.
Case/Case Relationship/Case Slot
The case is the functional
relationship of noun-phrases to the main action of sentences as expressed in
the verb. Every verb has a certain number of case slots. For example, the
verb `cut' presupposes someone cutting (the `agent' case slot), something
cut (the `object'), and something with which the act of cutting is performed
(the `instrument').
Central Processor
That part of a conventional digital computer
which performs the direct operations upon sequences of sysmbols stored as
sets of binary digits (bits). See also Architecture,
Compilation/Interpretation, Computation, Hardware,
Main Memory/Backup Memory, Parallel Distributed Processing,
Sequential Processing, Software.
Children
The relation of nodes to the node that immediately dominates
them in an isa hierarchy is that of children to a parent node. See
also Taxonomic Tree.
Cognitivism
In contrast to behaviourism this school of thought
claims that psychology should be concerned with a person's internal
representations of the world and with the internal or functional
organization of the mind. See also computationalism,
functionalism.
Command
Procedural programming languages such as POP-11
are composed of statements that can be interpreted as commands to the
computer to perform various operations -- for instance. adding two
numbers and assigning the result to a variable.
Commutative
A commutative relation in a semantic network is one
which is symmetric. That is, for any relation R, if xRy is true, then yRx is
also true. For example, if Frank is married to Hilda, then Hilda is also
married to Frank.
Competence Knowledge
An abstract description of the knowledge
characterizing a particular domain. See also
Performance Knowledge.
Competence/Performance
A person's linguistic competence is the
ability to produce and understand an indefinite number of novel sentences,
to detect ambiguities, and to recognize grammatical errors. Performance
refers to a person's use of language in particular situations. The
distinction is mirrored in natural language processing by grammars,
which are usually stored declarative representations of the rules
governing the well-formedness of sentences, and parsers, which are
programs that process utterances by reference to such rules.
Compilation/Interpretation
Two widely adopted methods for converting
the instructions of a program into binary machine code
instructions specifying basic operations on the machine's
central processor. Compilation involves converting a program as a
whole, whereas interpretation involves translating each instruction singly.
Compiled Representation
A representation of human knowledge in the
form of an automatic process that runs without calling on declarative
knowledge nor making demands on working memory.
Compositionality
The principle that the meaning (or
semantic value) of a sentence is a function of the meanings of its
constituents together with their mode of combination (just as the meaning of
a complex expression in propositional calculus is a function of the
truth-value of the atomic propositions together with the truth-functional
connectives by which they are combined together).
Computation
The manipulation of numbers or symbols according to fixed
rules. Usually applied to the operations of an automatic electronic
computer, but by extension to some processes performed by minds or
brains. See also Cognitivism.
Computationalism
The notion that the operation of the mind can be
explained entirely in terms of the formal, or functional, properties of a
computational system. See also Cognitivism, Eliminativism,
Functionalism, Materialism, Turing Test.
Computer
A machine for performing operations on numeric or symbolic
data. Nowdays the term usually refers to digital computers capable of
being instructed by stored programs. See also Algorithm,
Hardware, Software.
Concept
A concept is a stable mental representation of objects,
classes, properties, and relations. When we encounter a new object or event
for the first time, we draw upon our mental store of concepts in order to
identify it. One of the most important parts of the human learning process
is concept-formation, where, after a number of distinct experiences of the
`same' object or event, we acquire, by a process of induction, a
concept for it.
Conceptual Dependency Representation
A formalism for representing the
meanings of sentences, in a manner similar to that of a
semantic network, as structured configurations of basic
language-independent conceptual elements, or conceptual primitives.
See also Knowledge Representation, Semantics.
Conceptual Primitive
One of a small number of fundamental sorts of
action or event, such as the physical movement of an object (ptrans)
or the transfer of some mental content (mtrans). Used in conjunction
with scripts as a method of knowledge representation.
Conditional
Used to describe an instruction, or a sequence of
instructions, that is only to be executed when certain specified conditions
are met. For instance, the Automated Tourist Guide should only print
information about the river when the user has asked a question about the
river. Programming languages often introduce a conditional instruction by
means of the word if. The possibility that a program can carry out
different actions under different circumstances is called branching.
Conflict Resolution Strategy
A strategy used by a
production system interpreter for deciding which production rule is to
be fired if the heads of more than one rule match the contents of
working memory.
Connectionist Model/Neural Network
A system consisting of a network
of processing units each of which carries out a simple computation. The
units are all interconnected and knowledge is represented as patterns of
connection. See also Parallel Distributed Processing.
Constant
See Variable/Constant.
Data Object
Computer programs involve operations on different
kinds of symbolic entities, known as data objects or types. In POP-11
the data objects include numbers, words, booleans, and
lists.
Data Structure
A specific data object used as an object of
computation.
Database
A means of recording and accessing large amounts of
relatively simple items of data, such as bibliographical information. In
POP-11 the database package contains a set of built-in
procedures for accessing list-structured items.
Declarative/Procedural
A declarative representation contains
knowledge in the form of facts or images. In a procedural representation,
knowledge is stored as processes or procedures. Different programming
languages can be characterized as having a declarative or procedural style.
See also Procedural, Prolog, POP-11.
Declare
When a variable name is declared by means of the vars
instruction, POP-11 enters the word in its list of variable names and
allocates memory for the variable.
Deduction
A form of logical reasoning by which, from a given set of
facts (premises), certain consequences (conclusions) can be inferred.
For example, if I know that on every occasion on which it rains the
streets get wet, and that it is raining now, then I can deduce that the
streets are now wet.
Define
To specify the set of instructions associated with a
procedure, as used in computer programming.
Deictic Expression
An expression which picks out some individual,
time, or place relative to the context of utterance. For example, `I',
`you', `here', `these', `yesterday' have no intrinsic semantic value; the
reference of each on an occasion of speech is determined by the appropriate
component of the speech situation.
Demon
A type of procedure associated with a frame.
The two main types are if-needed and if-added demons.
Depth-First
A search strategy in which all the nodes on
one branch of the search tree are considered before backtracking
to try nodes on other branches. All the successors of a given node are
considered before any other nodes at the same level as the given node.
Digital/Analogue Computer
An analogue computer uses continuously
variable quantities, such as voltages, to represent similarly varying
enities in the domain. Digital computers operate upon discrete, coded
symbols.
Distinguished Symbol
In a generative grammar, the distinguished
symbol is the only non-terminal symbol which can appear at the
root of a parse-tree. It is conventionally `S' for `sentence'.
Distributed Processors
See Parallel Distributed Processing.
Distribution
Two syntactic items have the same distribution if they
can be substituted for each other in the same context within a sentence.
Domain
The area of enquiry and associated tasks to which an
artificial intelligence system is applied. (For instance, MYCIN has as its
problem domain the diagnosis of blood disorders.) See also
Domain-Independent/Domain-Specific Knowledge, Heuristic,
Knowledge-Based System.
Domain-Independent/Domain-Specific Knowledge
Many artificial
intelligence systems operate within a specific domain -- i.e., their
knowledge is mainly domain-specific. However, they may also contain
domain-independent knowledge, in the form of principles of common sense, or
general scientific laws common to many areas of expertise.
Dualism
A philosophical theory of mind associated with the
philosopher Descartes according to which human beings are constituted by two
distinct metaphysical substances or realms: Thought and Extension. See also
Functionalism, Materialism.
Dynamic Value
In game playing, the estimated value of a state,
arrived at by considering the results of a static evaluation function
applied to possible future states. These static values are propagated
back in time using, for instance, a minimax strategy.
Edge
An image contour along which intensity changes abruptly.
Edge Element
The smallest piece of edge detectable by examining
neighbouring pixels in an image.
Edge Map
A 2-D array recording the position of edge elements derived
from an image. Typically there is a one-to-one correspondence between array
cells and image pixels at the centre of edge elements.
Eliminativism
An approach to understanding the mind in which it is
claimed that terms like `beliefs' and `intentions' belong to an outmoded
`folk-psychology' and are not capable of scientific justification. See also
Functionalism, Intentional System/Intentional Stance,
Materialism.
ELIZA
A program which can be taught to respond to sentences typed by
a user in a way that can give the illusion of understanding them, but which
in fact relies on pattern matching and substitution to produce
answers based on a pre determined set of stock phrases. ELIZA was originally
developed by Joseph Weizenbaum in the mid-1960s, and now serves as a
demonstration of the technique and as a warning against taking the responses
of a computer at face value.
Embed
When a syntactic constituent has another of the same category
nested within it as one of its constituents, the nested constituent is said
to be embedded within the larger one.
Evaluate
To perform an operation on given arguments producing a
canonical (or simplified) result. For example, given an appropriate
definition for the procedure maximum, the expression
maximum(5,7) will evaluate to 7.
Expert System
A computer program that incorporates the
rule-governed expertise of human experts such as doctors, engineers, or
economists within a restricted problem domain. See also
Heuristic, Knowledge-Based System.
Expression
In a computer program, a combination of
data objects linked by operators.
Feature
See Attribute/Feature/Property.
Fire
(Pertaining to production rules.) See Trigger.
Flow of Control
The sequential behaviour of a program, as it
carries out the processing of input.
Formal semantics
A theory of meaning based on the analysis of a
natural language as though it were a formal language, such as the
propositional and predicate calculi. Typically, it is
truth-conditional and compositional.
Formalism
A set of symbols and a collection of systematic rules
governing their uses and interrelations. See also
Knowledge Representation, Logic.
Formula
In predicate logic a formula consists either of a
predicate applied to one or more aguments, or a combination of
such `atomic' formulae linked by the connectives `not', `or', and `and'. See
also Axiom/Theorem.
Frame
Frame systems are a particular kind of
knowledge representation formalism, consisting of structured
descriptions of stereotyped objects or situations, with slots filled
in with particular values. See also Demon, Script.
Functionalism
A philosphical view of mind according to which mental
processes are characterized in terms of their abstract functional (opr
computational) relationships to one another, and to sensory inputs and motor
outputs. See also Cognitivsm, Computationalism,
Eliminativism, Turing Test.
Generate
A formal set of rules which explicitly defines the
grammatical sentences of a language is said to generate that language. Such
a set of rules constitutes a generative grammar.
Generic
A generic term is one which picks out a class of individuals,
or the prototype of the individual, rather than the individual itself. For
example, in the sentence ``The wolf has disappeared from northern Europe,''
we are referring to the genus rather than to a particular wolf.
Grammar
A grammar may be informally thought of the set of words and
category symbols in a language, together with the rules governing their
combination into sentences. It is, in human terms, what speakers know of
their language independent of their ability to use that knowledge. See also
Competence Knowledge, Competence/Performance, Syntax.
Grammatical
A grammatical sentence is one which is well-formed with
respect to the rules (`grammar') for generating the sentences of the
language. See also Syntax.
Grey-Level
A numerical measure of the intensity of a single pixel.
The number is typically an integer between 0 and 63 or 0 and 255, where 0
signifies minimum intensity and 63 or 255 signifies maximum intensity.
Hardware
The physical machinery of a computer and its
associated devices. See also Algorithm, Central Processor,
Computer, Program, Software,
Visual Display Unit (VDU).
Heuristic
A rule of thumb which may help in solving a problem, but is
not guaranteed to find a solution. A method for cutting down useless
searching amongst possibilities. Heuristics provide much of the knowledge
contained in expert systems. Constrasts with algorithmic
procedures. See also Knowledge-Based Systems.
Heuristic Search
Any search strategy which makes use of
heuristics to suggest the best nodes to consider at each stage
of the search.
Induction
A form of reasoning by which we derive general principles
from particular facts or instances sharing common properties. For example,
if all ravens of which we have had experience are black, then we might
inductively reason that all ravens are black.
Information Processing System
Generally speaking, any system which
operates on symbolic data given as its input so as to generate an
appropriate output. See also Computer, Program.
Inherit
See Property Inheritance.
Input/Output
(a) The values that are passed into a procedure, and
which are returned as the results of the operation of a procedure.
(b) In general, those aspects of a computer system which involve the
transmission of data into the system, and the communication of results back
to the outside world. See also Argument/Result,
Calling a Procedure.
Intentional System/Intentional Stance
Some philosophers claim that
intentionality should be understood instrumentally as a feature of certain
kinds of systems; these would include intelligent computers as well as
people. It is necessary use terms such as `intending', `believing',
`wanting' in order to explain the behaviour of such systems. Adopting this
`intentional stance' does not commit one to believing that there really are
intentional states as such. See also Cognitivsm,
Computationalism, Eliminativism, Functionalism,
Turing Test.
Intentionality
The quality of `aboutness', or `world-directedness',
that has been claimed to be a distinguishing feature of mental processes as
opposed to physical ones. It has been proposed that `intelligent'
computer performances possess only a secondary, or derivative,
intentionality: this has been used as an attack on the pretensions of
artificial intelligence to explain the human mind. See also
Computationalism, Functionalism,
Intentional System/Intentional Stance, Turing test.
Internal Representation
A symbolic representation in a computer
of the states and processes of problem solving, search, etc. See also
Cognitivsm, Knowledge Representation.
Interpretation
See Compilation/Interpretation.
Introspection
A process of inward attention or reflection, so as
to examine the contents of the mind. Introspectionism, as a method in
psychology, is contrasted with, and was supplanted by, behaviourism.
isa Hierarchy
A series of nodes joined by links of the form `X isa Y'
in semantic networks, which allow efficient forms of inference.
See also Knowledge Representation, Property Inheritance,
Structural Description.
Keyword
In a program based on pattern matching, a keyword is a
word in the input sentence which determines the actions to be taken. In a
simple program such as ELIZA, recognition of a keyword may completely
determine the response of the program; in more sophisticated programs the
keyword may determine the context in which further analysis is undertaken.
Knowledge Representation
The term used in artificial intelligence to
cover the study of formalisms which model human forms of knowledge.
See also Frames, Predicate Calculus/Predicate Logic,
Production Systems, Scripts, Semantic Networks.
Knowledge-Based System
A computer program which represents
relatively complex kninds of information, often as rules governing
judgements in particular problem domains. See also Database,
Expert System, Knowledge Representation.
Lexical Category
Every word in the language belongs to some lexical
category, that is, instances some part of speech. Thus, `dog', `brain', and
`syllogism', for example, all belong to the lexical category `noun'.
Lexicon
The lexicon of a language is dictionary of all the words in
the language, and may contain many types of information about each word, for
example, what part of speech it is (its lexical category), and what
its distributional properties are.
Link
See Arc/Link/Pointer.
LISP
A programming language that uses the list as its primary
data object. The most widely used language for artificial
intelligence.
List
A special kind of data object, found in POP-11
and other artificial intelligence languages. A list consists of a sequence
of zero or more elements, each of which may be any kind of data object
(including another list).
Local/Global
In POP-11 procedures, variables can be
defined as local to a given procedure. These variables lose their
values outside the context of the procedure itself. Global variables,
however, retain their assigned values after the procedure
terminates.
Logic
The study of formal principles of rasoning. The two fundamental
systems of logic are propositional calculus and
predicate calculus, although many extensions to these systems exist.
Logic is used extensively in computer science and artificial intelligence.
See also Axiom/Theorem, Rules of Inference.
Main Memory/Backup Memory
The part of the computer's circuitry
which provides instructions and data to be fed into the
central processor, and which stores the output of the latter's
operation, is known as the main memory. By contrast, the backup memory, such
as disc or tape, contains information in a form which can be transported
from one machine to another.
Materialism
A general philosophical view that only physical processes
exist. The term is also applied to philosophical theories of mind which
claim that mental states are identical with brain states. See also
Dualism, Eliminativism, Functionalism.
Microworld
A representation of some well-defined domain, such
as Newtonian physics, such that there is a simple mapping between the rules
and structures of the microworld and those of the domain.
Minimax
A strategy to determine good moves in computer game playing
by choosing the move that minimizes the loss the player can expect to incur,
under the assumption that the player's opponent is adopting the same
strategy (and hence trying to cause the maximum loss). The possible gain or
loss for a sequence of moves is usually assessed using a
static evaluation function applied to the state of play at the end of
the sequence. See also Dynamic Value.
Model
In vision, an internal description of the structure of an
object, whether in terms of the structure of image features in a
2-dimensional image or solids in 3-dimensional space.
Naive Physics
People's commonsense knowledge about the behaviour of
physical objects in the world -- for example, that a liquid spreads out when
it pours onto a flat surface.
Neural Network
See Connectionist Model.
Node
A single state in a state-space representation. In
search, a junction point in the search tree, representing a
state which may need to be considered either as a possible solution to
a problem or en route to reaching such a solution.
Non-terminal Symbol
In a generative grammar, a non-terminal symbol
labels a syntactic or lexical category, and is used in the formulation of
the rules of the grammar. Examples of non-terminals are Prep, VP, Det,
Pronoun, NP.
Operator
(a) A symbol to denote an operation, such as multiplication
or comparison, which is applied to numbers or other data objects in a
computer program.
(b) A discrete atomic mental operation that transforms one knowledge state,
or state of working memory, into another. See also
Problem Behaviour Graph.
Operator Schema
A knowledge representation formalism similar to
a frame representing the type, preconditions, and actions of an
operation. Used in artificial intelligence planning and problem solving
systems.
Parallel Distributed Processing
Computing by means of a network of
simple computational elements, working in parallel, each of which influences
the other elements and is influenced by them. See also
Connectionist Model, Sequential Processing.
Parser
A program which, given as input a grammar, a
lexicon, and a word-string, will output, if the string is a
well-formed sentence, a structural description of the sentence. Otherwise,
the parser will reject the string. The structural description is often
represented as a parse-tree, also called a phrase-marker,
a diagrammatic representation of the sentence's constituent structure.
Parse-Tree
See Parser.
Pattern Matching
A programming method for matching a list of words
(or other symbols) against a template, to discover whether particular words
or combinations of words occur in the list. Pattern matching is built into
POP-11. See also Wild Card.
Performance
See Competence/Performance.
Performance Knowledge
The knowledge posessed by a particular person
which determines what or how they will perform on a given task See also
Competence Knowledge.
Phrase-Marker
A diagrammatic representation, in the form of an
inverted schematic tree, of a sentence's constituent structure. See also
Parser.
Phrase-Structure Rule
A rule which specifies the internal structure
of a syntactic category. Rules are generally written in the following
manner: X [IMAGE ] Y Z, where X, Y, and Z are syntactic category
symbols, and Y and Z can replace X in its context within a sentence. A set
of such phrase-structure rules, where at least one of the rules has the
distinguished symbol on the left-hand side of the arrow, constitutes a
phrase-structure grammar.
Pixel
A tiny patch of an image. Images are normally broken down into
a rectangular lattice of square pixels like tiles on a wall, although
hexagonal tessellations have also been used.
Plan
A routinized program for behaviour that is triggered when
certain specific environmenal conditions are satisfied.
Planning
The selection of a sequence of operations in order to bring
about one or more goals. An important area of enquiry of artificial
intelligence. See also Problem Behaviour Graph,
Problem Reduction.
Pointer
See Arc/Link/Pointer.
POP-11
A programming language for artificial intelligence teaching
and research.
Pragmatics
Those aspects of the study of language that pertain to the
identity and intentions of the speaker and hearer, and the context in which
speech takes place. The context is sometimes most narrowly regarded as the
body of world knowledge to which speakers and hearers have access in
generating and interpreting speech. Pragmatics belongs to the study of
linguistic performance. See also semantics, speech act,
syntax.
Predicate
In logic, a function whose value is is either `true',
or `false'. It specifies a relatioship between arguments. See also
Axiom/Theorem, Formula.
Predicate Calculus/Predicate Logic
A system of logic in which
one can formalize inferences by representing relationships between
predicates such as `...is happy' and `...is wise'. Predicate
calculus incorporates, but extends, propositional calculus. See also
Axiom/Theorem, Logic, Quantifier/Quantification.
Problem Behaviour Graph
A schematic representation of a subject's
problem space, in the form of labeled boxes denoting knowledge states,
and directed links representing the mental operations that transform
each such state into a new state. See also Knowledge Representation,
Planning.
Problem Reduction
A method of solving problems by splitting them into
more manageable parts.
Problem Space
A set of knowledge states, including a start-start and
a goal-state, and a set of admissible problem-specific operations which can
be performed on these states. The problem space can be represented
schematically by a problem behaviour graph.
Procedural
See Declarative/Procedural.
Procedural Programming Language
A programming language such as
POP-11 consisting procedure definitions together with sequences
of commands for invoking them.
Procedure
A sequence of instructions in a programming language
designed to carry out a coherent operation or set of operations. Procedures
often operate on data, passed to them through arguments, and return
results, which may be used by other parts of the program. Procedures are
first defined, to establish the actions they take, then called
to carry out those actions.
Production System
A knowledge representation formalism
consisting of collections of condition-action rules (called production
rules), a database which is modified in accordance with the rules, and
a production system interpreter which controls the operation of the
rules. See also Conflict Resolution Strategy.
Production System Interpreter
The `control mechanism' of a
production system, determining the order in which production rules are
fired.
Program
A set of abstract instructions or procedures for a
computer to perform. When the instuctions are written in a high-level
language like POP-11, they have to be converted, either by
compilation or interpretation, into basic operations that the
computer can perform. See also Algorithm, Hardware,
Software.
Prolog
A programming language with a strongly declarative
style, which is based on predicate calculus. See also
Resolution.
Property
See Attribute/Feature/Property.
Property Inheritance
A process by which features attached to
parent nodes (more general concepts or inclusive classes) in an
isa hierarchy are passed down to (`inherited by') their children (more
specific concepts or sub-classes) in such cases where no properties or
attributes of the same type are associated explicitly with the children
nodes. For example, if we know that birds can fly and that canaries are
birds, then we can infer that canaries too can fly.
Propositional Calculus
A simple but powerful system of logic
for representing implications between logically compound sentences, such as
`either A or not B', `if B then A', etc. The system will prove, for example,
that each of these two sentences implies the other. See also
Predicate Calculus/Predicate Logic, Rules of Inference.
Protocol Analysis
A method of studying subjects' mental processes in
the performance of tasks by recording their spontaneous `thinking aloud' and
subsequently segmenting the running commentary into the discrete atomic
mental operations that the subjects have used in the accomplishment of the
tasks. See also Operator, Problem Space.
Quantifier/Quantification
Quantifiers play an important role in
predicate calculus logic. There are two kinds of quantifier: the
universal quantifier (as in `all dogs are mammals') and the singular or
existential quantifier (as in `some mammals are dogs').
Recursion
Self-reference, particularly in the form of a
procedure containing, within its body, a call to iteslf. A recursive
phrase-structure rule is one in which the symbol on the left-hand
side of the arrow also appears on the right-hand side -- for example,
NP [IMAGE ] NP Prep NP, which says that a noun-phrase can be made
up of two embedded noun-phrases joined by a preposition.
Region
An area of an image which is homogeneous in intensity or
texture.
Region Map
A 2-D array representing a segmentation of an input image
into regions. The array is typically the same size as the input image, so
that there is a one-to-one correspondence between array cells and image
pixels. Array cells belonging to the same region contain a label which is
unique to that region.
Repair Heuristic
A general-purpose heuristic used to effect a
repair.
Repair Theory
An analysis of errors, based on the assumption that a
learner who reaches as impasse, as a result of following a wrong method,
will devise a repair to be able to continue solving the problem.
Repeat Loop
A structure in a program which allows an instruction, or
a sequence of instructions, to be carried out repeatedly. The repetition may
continue indefinitely, or for a fixed number of times, or may end when some
condition specified in the program becomes true. See also Until Loop,
While Loop.
Resolution
A popular technique used in automated theorem proving
systems. A version of resolution underlies the logic programming
language Prolog.
Restriction Procedure
Matching of variables in lists can
be constrained, in POP-11, by a restriction procedure, which specifies
some property or feature a variable must have.
Result
See Argument/Result, Input/Output.
Root
The node of a search tree from which all other nodes
can be reached by moving down the branches. Often represents the starting
state of a problem or game.
Rule Body
The `action' part of a production rule, which specifies
what actions are to be performed if the rule is fired (i.e., there is a
successful match of the rule head against the contents of the
working memory). Most typically, these are actions which change the
contents of the working memory by adding and removing facts.
See also Conflict Resolution Strategy, Production System.
Rule Head
The `condition' part of a production rule (in the
POP-11 production system formalism, the first line of a production
rule) consisting of one or more facts which the
production system interpreter attempts to match against the contents
of the working memory. See also Conflict Resolution Strategy,
Rule Body.
Rule-by-Rule Principle
Maintains that the meaning of a sentence can
be compositionally built up by semantic rules in tandem with the syntactic
parse. See also Formal Semantics.
Rulebase
The set of rules in a production system. See also
Rule Body, Rule Head.
Rules of Inference
Any system of logical deduction uses general
rules of inference to allow logical formulae to be derived from
others. For instance, a rule of inference in many systems of
propositional calculus is modus ponens, which, given the
propositions `if P then Q' and `P', allows one to infer `Q'.
Script
A structure rather like a frame which represents
situations or events, and which is used as a knowledge representation
formalism. It is particularly suited to natural language understanding
systems. See also Conceptual Primitive.
Search
A technique of problem solving in which possible solutions are
explored in a systematic way, in effect by asking a series of ``What if
...?'' questions. Asking each a question corresponds to trying out a
possible state of the domain in which the program operates to
see if it leads to a solution of the problem.
Search Strategy
Any way of organizing the questions asked in problem
solving by search, so as either to find a solution as rapidly as
possible or to find a particularly good solution when more than one
solution is possible.
Search Tree
The exploration of possibilities in problem solving by
search generally leads to a branching structure, in which each
state may give rise to a new set of states (sometimes called successor
states) and each of these may give rise to successor states of its own,
and so on. This structure is not usually represented in full in
computer memory, but is implicit in the problem itself. Finding a
solution involves systematically traversing this tree.
Semantic Network
A knowledge representation formalism which
describes objects and their relationships in terms of a network consisting
of labelled arcs and nodes.
Semantic Value
In formal semantics, the semantic value of an
expression is the individual(s) or state(s) in the world to which the
expression refers, or which make that expression true.
Semantics
The study of meaning in language. See also
formal semantics, pragmatics, syntax.
Sequential Processing
A computer architecture in which a
central processor carries out a series of operations in sequence.
See also Computation, Parallel Distributed Processing.
Set-Value Variable
In POP-11 a set-value variable in a pattern
matches an equivalent element in the list being matched. If there is a
correct match, then the item in the list is assigned to the variable.
See also Use-Value Variable, Wild Card.
Slot
An element in a frame, which can be filled with a value.
For instance a `chair' frame may have a `number of legs' slot, perhaps
filled with the default value of 4.
Software
A program for a computer, together with the data
upon which the program operates. See also algorithm, hardware.
Speech Act
The theory of speech acts views natural language
utterances as actions -- for example, questions, orders, statements -- on
the part of speakers. Speech act theory belongs to the field of
pragmatics.
State
A description of a possible arrangement of a domain.
For instance, in game playing, a state may correspond to a position of
pieces on the board; in route finding, a state may correspond to arrival at
a specified place at a specified time; in planning, a state may correspond to
an arrangement in space of some objects to be manipulated.
State-Space
The set of all the states which could possibly need
to be considered in the course of solving a particular problem.
Static Evaluation Function
In computer game playing, any method of
determining the advantage to the player of a given state of play, without
looking ahead at possible future moves.
Structural Description
A description of the formal constituents,
features, and relationships that exhaustively account for the
structural integrity and cohesion of an object. A structural description of,
for example, an arch will include such terms as `block', `supported by', and
`abutting'. See also Knowledge Representation, Semantic Network.
Substitution
In a program such as ELIZA, using part of the
user's input sentence to produce response. In general, replacing part of a
string of words or other symbols by different symbols, depending on context.
Successive Refinement
A method of software design
characteristic of artificial intelligence program development, which
involves starting with a relatively crude working program and gradually
increasing its complexity and sophistication.
Syntactic Category
A grammatical class determined on the basis of the
syntactic role it fills in the construction of sentences, used in the formal
description of language. It is labelled by a category symbol, such as NP,
VP, PrepP. See also Phrase-Structure Rule, Syntax.
Syntactically Well-Formed
See Grammatical.
Syntax
(a) In programming, the rules by which the words in a program
are combined to form commands to a computer.
(b) For natural languages, the study of the rules governing the way in which
words and phrases are combined to form sentences in a language. The word is
also sometimes used to refer to the rules themselves; thus one may speak of
`English syntax', `the syntax of German', and so on, to mean much the same
thing as `English grammar', `the grammar of German', and so forth. See also
Grammar.
Taxonomic Tree
See isa Hierarchy.
Template
In vision, a stored sub-image with which to compare incoming
images.
Terminal Symbol
Terminal symbols are the output of the syntactic
component of a grammar, and are usually identified with the words and
grammatically meaningful sub-parts (`morphemes') of words. In a
phrase-structure grammar, they will only ever appear on the right-hand
side of lexical rules. See Lexical Category, Lexicon,
Phrase-Structure Rule, Syntax.
Theorem
See Axiom/Theorem.
Transitive
A relationship which, for any relation R, if xRy and yRz,
then xRz. For example, if a canary is a bird and a bird is an animal, then a
canary is an animal.
Trigger
A production rule is said to be triggered if there is a
successful match of the rule head against the facts in
working memory. See also Conflict Resolution Strategy,
Production System.
Truth-Conditionality
The assumption that meaning can be defined in
terms of the conditions in the real world under which a sentence may be used
to make a true statement. See also Formal Semantics.
Turing Test
A hypothetical test for computer intelligence,
proposed by Alan Turing in 1950, involving a computer program
generating a conversation which could not be distinguished from that of a
real human.
Until Loop
A structure in a computer program which allows a
sequence of instructions to be repeated until some condition becomes true.
See also Repeat Loop, While Loop.
Use-Value Variable
In POP-11 when a use-value variable occurs
in a list, the value of the variable is inserted at that place in the list.
See also Set-Value Variable, Wild Card.
User-Defined
See Built-In/User-Defined.
Value
The actual data which have been stored in a section of
computer memory and which are associated with a variable.
The value can change as new assignments to the variable are made.
Variable/Constant
In many programming languages, there is a class of
terms which do not have a fixed value, but to which values can be
assigned. These are called variables. Constants, on the other hand,
are terms whose values are fixed. See also Local/Global.
Visual Display Unit (VDU)
A screen, usually like a TV set, that
displays text or pictures representing the output of a program
while it is running on a computer. See also Hardware,
Input/Output.
von Neumann Architecture
The `classic' design for
computer hardware, named afer the mathematician John von Neumann, who
formulated the main elements of the standard, sequential, stored program,
computer architecture in the late 1940s. Non-von Neumann architectures
are now being developed that exploit various forms of parallel
or distributed processing. See also central processor,
computation, connectionist model,
parallel distributed processing.
While Loop
A structure in a computer program that allows a
sequence of instructions to be repeated while some condition remains true.
See also Repeat Loop, Until Loop.
Wild Card
In pattern matching, an element in a pattern which
can match any element in the equivalent section of the list being
matched. See also Set-Value Variable, Use-Value Variable.
Word
A particular kind of data object. In POP-11,
a word identified by sequences of symbols within double quotation marks,
such as "fred", "abc123". Each word used by a program is stored
in an internal dictionary.
Word Concept
The stored mental representation of a word.
Working Memory
Sometimes also called `short term memory' by
psychologists, it is a mental workspace consisting of a small set of data
items representing the current state of knowledge of a system at any stage
in the performance of a task, and which is transformed into a new set, in
humans by the application of a discrete mental operation (operator),
and in production systems on the firing of a new production rule. See also
Production System.
![[IMAGE ]](img1.gif){kind=link}