PROLOG EXPERT SYSTEM BUILDING TOOL
          Contributed by Steven Salvini, Heriot-Watt University
                   Received on the 30th of August 1988
                   Shelved on the 3rd of October 1988


PROTEST EXPERT SYSTEM

This is a  cut-down version of the expert system building tool produced
by Steven Salvini as part of his M.Sc. dissertation in 1986.

It comes as the following files:

    SALVINI.PRE         - this file.

    COMPAT.PL           - source.
    TERM.PL
    WRITEF.PL
    OPS.PL
    EXPLORE.PL
    MAIN.PL
    USER_ANSWER.PL
    SHOW_ANSWER.PL
    HISTORY.PL
    UTILITIES.PL

    DENTAL.PL           - knowledge bases.
    WEE.PL

    MANUAL.TXT          - the user manual.

    XREF.TXT            - these are the original files, before
    XREF1.TXT             I modified them.
    DENTAL.OLD
    EXPLORE.OLD
    HISTORY.OLD
    LOADER.OLD
    MAIN.OLD
    OPS.OLD
    SHOW_ANSWER.OLD
    TERM.OLD
    TERMCAP.OLD
    USER_ANSWER.OLD
    UTILITIES.OLD
    WEE.OLD
    WRITEF.OLD
    READ.ME


To use Protest, consult LOADER.PL, and call 'go'.

Protest is a  tool for building prototype expert  systems. Its knowledge
bases  are written  in IF-THEN  rules whose  conditions and  conclusions
consist of object/attribute/value triples:

    rule1:
        if
                pain/duration/constant
            and
                pain/localisation/accurate
            and
                pain/'initiated by'/'gentle touch'
            and
                pain/character/'sudden onset'
            and
                pain/character/'intense & throbbing'
            and
                gingivae/swelling/present
            and
                gingivae/erythema/present
            and
                miscellaneous/'local lymphadenopathy'/present
            and
                pain/'initiated by'/'clenching, biting or chewing'
            and
                tooth/mobility/present
        then
            diagnosis/group/abscess
        with
            certainty(0.9).

Conclusions may contain certainty factors.

Inference can be either backward chaining or mixed. The former is normal
backward  chaining, giving  a single  firm  conclusion if  one has  been
deduced or  else as  a list of  possible conclusions  in order  of their
certainties.

Mixed mode  is a "mixed" inference  strategy which first prompts  you to
enter your initial findings and then tries  to deduce one or more of the
goals. If a firm  conclusion is not reached, it then  tries to solve the
problem by backward-chaining.


CHECKED ON EDINBURGH-COMPATIBLE (POPLOG) PROLOG : yes.

    occasionally, some messages seem to  go missing, possibly because of
    an incompatibility in terminals.

PORTABILITY :

    the terminal-control predicates are terminal-specific.


INTERNAL DOCUMENTATION :

    Each important predicate  comes with a brief  comment describing its
    purpose. There is a user manual.