view release on metacpan or  search on metacpan

libmegahal.c  view on Meta::CPAN

 *
 *  You should have received a copy of the GNU General Public License along
 *  with this program; if not, write to the Free Software Foundation, Inc.,
 *  675 Mass Ave, Cambridge, MA 02139, USA.
 */

/* Modified version
Alexandr Ciornii
  Win32 (MSVC & gcc), FreeBSD, Mac Darwin compatibility
  const'ing function parameters
Craig Andrews
  void megahal_learn
*/

/*===========================================================================*/

/*
 *		$Id: megahal.c,v 1.6 2002/10/16 04:32:53 davidw Exp $
 *
 *		File:			megahal.c
 *
 *		Program:		MegaHAL
 *
 *		Purpose:		To simulate a natural language conversation with a psychotic
 *						computer.  This is achieved by learning from the user's
 *						input using a third-order Markov model on the word level.
 *						Words are considered to be sequences of characters separated
 *						by whitespace and punctuation.  Replies are generated
 *						randomly based on a keyword, and they are scored using
 *						measures of surprise.
 *
 *		Author:		Mr. Jason L. Hutchens (http://www.amristar.com.au/~hutch/)
 *
 *		WWW:		http://megahal.sourceforge.net
 *
 *		Compilation Notes
 *		=================
 *
 *		When compiling, be sure to link with the maths library so that the
 *		log() function can be found.
 *
 *		On the Macintosh, add the library SpeechLib to your project.  It is
 *		very important that you set the attributes to Import Weak.  You can
 *		do this by selecting the lib and then use Project Inspector from the
 *		Window menu.
 *
 *		CREDITS
 *		=======
 *
 *		Amiga (AmigaOS)
 *		---------------
 *		Dag Agren (dagren@ra.abo.fi)
 *
 *		DEC (OSF)
 *		---------
 *		Jason Hutchens (hutch@ciips.ee.uwa.edu.au)
 *
 *		Macintosh
 *		---------
 *		Paul Baxter (pbaxter@assistivetech.com)
 *		Doug Turner (dturner@best.com)
 *
 *		PC (Linux)
 *		----------
 *		Jason Hutchens (hutch@ciips.ee.uwa.edu.au)
 *
 *		PC (OS/2)
 *		---------
 *		Bjorn Karlowsky (?)
 *
 *		PC (Windows 3.11)
 *		-----------------
 *		Jim Crawford (pfister_@hotmail.com)
 *
 *		PC (Windows '95)
 *		----------------
 *		Jason Hutchens (hutch@ciips.ee.uwa.edu.au)
 *
 *		PPC (Linux)
 *		-----------
 *		Lucas Vergnettes (Lucasv@sdf.lonestar.org)
 *
 *		SGI (Irix)
 *		----------
 *		Jason Hutchens (hutch@ciips.ee.uwa.edu.au)
 *
 *		Sun (SunOS)
 *		-----------
 *		Jason Hutchens (hutch@ciips.ee.uwa.edu.au)
 */

/*===========================================================================*/

#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#ifndef _MSC_VER
#include <unistd.h>
//#include <getopt.h>
#endif
#if !defined(AMIGA) && !defined(__mac_os) && !defined(__FreeBSD__) && !defined(__APPLE__)
// FreeBSD malloc.h is empty and gives error
// Tested on FreeBSD 5.4
#include <malloc.h>
#endif
#include <string.h>
#include <signal.h>
#include <math.h>
#include <time.h>
#include <ctype.h>
#if defined(__mac_os)
#include <types.h>
#include <Speech.h>
#else
#include <sys/types.h>
#endif
#include "megahal.h"
#if defined(DEBUG)
#include "debug.h"
#endif

libmegahal.c  view on Meta::CPAN

/*---------------------------------------------------------------------------*/
/*
 *    Function:   Generate_Reply
 *
 *    Purpose:    Take a string of user input and return a string of output
 *                which may vaguely be construed as containing a reply to
 *                whatever is in the input string.
 */
char *generate_reply(MODEL *model, DICTIONARY *words)
{
    static DICTIONARY *dummy=NULL;
    DICTIONARY *replywords;
    DICTIONARY *keywords;
    float surprise;
    float max_surprise;
    char *output;
    static char *output_none=NULL;
    int count;
    int basetime;
    int timeout = TIMEOUT;

    /*
     *		Create an array of keywords from the words in the user's input
     */
    keywords=make_keywords(model, words);

    /*
     *		Make sure some sort of reply exists
     */
    if(output_none==NULL) {
	output_none=malloc(40);
	if(output_none!=NULL)
	    strcpy(output_none, "I don't know enough to answer you yet!");
    }
    output=output_none;
    if(dummy == NULL) dummy = new_dictionary();
    replywords = reply(model, dummy);
    if(dissimilar(words, replywords) == TRUE) output = make_output(replywords);

    /*
     *		Loop for the specified waiting period, generating and evaluating
     *		replies
     */
    max_surprise=(float)-1.0;
    count=0;
    basetime=time(NULL);
/*     progress("Generating reply", 0, 1);  */
    do {
	replywords=reply(model, keywords);
	surprise=evaluate_reply(model, keywords, replywords);
	++count;
	if((surprise>max_surprise)&&(dissimilar(words, replywords)==TRUE)) {
	    max_surprise=surprise;
	    output=make_output(replywords);
	}
/*  	progress(NULL, (time(NULL)-basetime),timeout); */
    } while((time(NULL)-basetime)<timeout);
    progress(NULL, 1, 1);

    /*
     *		Return the best answer we generated
     */
    return(output);
}

/*---------------------------------------------------------------------------*/

/*
 *		Function:	Dissimilar
 *
 *		Purpose:		Return TRUE or FALSE depending on whether the dictionaries
 *						are the same or not.
 */
bool dissimilar(DICTIONARY *words1, DICTIONARY *words2)
{
    register int i;

    if(words1->size!=words2->size) return(TRUE);
    for(i=0; i<words1->size; ++i)
	if(wordcmp(words1->entry[i], words2->entry[i])!=0) return(TRUE);
    return(FALSE);
}

/*---------------------------------------------------------------------------*/

/*
 *		Function:	Make_Keywords
 *
 *		Purpose:		Put all the interesting words from the user's input into
 *						a keywords dictionary, which will be used when generating
 *						a reply.
 */
DICTIONARY *make_keywords(MODEL *model, DICTIONARY *words)
{
    static DICTIONARY *keys=NULL;
    register int i;
    register int j;
    int c;

    if(keys==NULL) keys=new_dictionary();
    for(i=0; i<keys->size; ++i) free(keys->entry[i].word);
    free_dictionary(keys);

    for(i=0; i<words->size; ++i) {
	/*
	 *		Find the symbol ID of the word.  If it doesn't exist in
	 *		the model, or if it begins with a non-alphanumeric
	 *		character, or if it is in the exclusion array, then
	 *		skip over it.
	 */
	c=0;
	for(j=0; j<swp->size; ++j)
	    if(wordcmp(swp->from[j], words->entry[i])==0) {
		add_key(model, keys, swp->to[j]);
		++c;
	    }
	if(c==0) add_key(model, keys, words->entry[i]);
    }

    if(keys->size>0) for(i=0; i<words->size; ++i) {