See ../guile/README for an intro on running the guile shell needed
to run the demos here.
The first three examples provide a basic introduction to basic pattern matching.
- simple.scm: A basic introduction to pattern matching.
- satisfcation.scm: An example of using the SatisfactionLink
- glob.scm: Matching multiple atoms at once.
- choice.scm: Using the ChoiceLink to explore alternatives.
The pattern matcher can be used to trigger side-effects, when a pattern is matched. This includes the execution of arbitrary code, both as "black-box" code, as well as "clear-box" Atomese.
- gsn.scm: Calling arbitrary functions upon a match.
- gpn.scm: Calling arbitrary functions to decide a match.
- sequence.scm: Using GPN's to execute a sequence of tasks.
- condition.scm: Combining GPN's and GSN's to make an action taken depend on a precondition.
The pattern matcher is a kind of graph query-engine; it is looking for the presence (or absence) of certain graphs in the atomspace. This can be understood to be a form of "intuitionistic logic", or "constructivist logic".
- presence.scm: various ways of testing for the presence of an atom in the AtomSpace.
- absent.scm: Use the StateLink to set state; use AbsentLink to check for the absence of atoms in the AtomSpace.
The above examples should make clear that the pattern matcher implements a kind-of programming langauge, refered to as "atomese". Like any good programming language, it steals ideas. One of these is the idea of defining things. Another is a fairly complete type system, providing basic type constructors (although it currently stops short of providing dependent types).
- define.scm: using DefineLinks to create patterns out of parts.
- type-signature.scm: using signatures and type constructors to refine the search.
The pattern matcher can be used to implement a rule engine. In order for rule engines to run efficiently, they need to implement the Rete algorithm, or something similar. The pattern-recognizer, or DualLink, does this: it is used to find the rules that can be applied to a given graph.
- recognizer.scm: using DualLink to implement an AIML-like system.
Some simpler applications showing some things one can do, and how to do them.
- fsm-basic.scm: A simple Deterministic Finite State Machine demo.
- fsm-full.scm: A generic deterministic finite-state-machine constructor.
- fsm-mealy.scm: A generic Mealy machine constructor.
- markov-chain.scm: A Markov chain (probabilistic FSM) based on fsm-full.
Examples that demonstrate some of the inner workings of the pattern matcher.
- virtual.scm: Using virtual links, and the resulting combinatorial explosion.
Unfinished examples:
- deduction-engine.scm: How to create a ProLog-like reasoning engine in Atomese.