source: trunk/indexers/mgpp/lib/lovinstem.cpp@ 9612

/**************************************************************************
 *
 * lovinstem.c -- Stemming code
 * Copyright (C) 1994  Linh Huynh ([email protected])
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * 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.
 *
 **************************************************************************/

#include "sysfuncs.h"


/*****************************************************************************
 *
 *   Transformational Rules used in Recoding Stem Terminations
 *
 *   Parameter in functions:
 *       ch : the second last character of the old ending
 *            (after removing one of double consonants) 
 *
 ****************************************************************************/


/************ Conditional rules associated with transformations **************/

/*static int aio ();
static int s ();
static int pt ();
static int m ();
static int n ();*/

static int 
aio (char ch)           /* Rule number 9     */
{
  return ((ch != 'a') && (ch != 'i') && (ch != 'o'));
}

static int 
s (char ch)             /* Rule Number 24    */
{
  return (ch != 's');
}

static int 
pt (char ch)                /* Rule number 28    */
{
  return ((ch != 'p') && (ch != 't'));
}

static int 
m (char ch)             /* Rule number 30    */
{
  return (ch != 'm');
}

static int 
n (char ch)             /* Rule number 32    */
{
  return (ch != 'n');
}


/**************************************************************************
 *
 *  Context-sensitive rules associated with certain endings.
 *
 *  The notations of the rules are the same as in Lovins' paper
 *  except that rule A is replaced by a NULL in the data structure.
 *
 *  Function parameters:
 *      - stem_length: possible stem length
 *      - end        : pointer points to the end of the possible stem.
 *
 *************************************************************************/


/********* Context-sensitive rule function declarations ******************/

/*static int B ();
static int C ();
static int D ();
static int E ();
static int F ();
static int G ();
static int H ();
static int I ();
static int J ();
static int K ();
static int L ();
static int M ();
static int N ();
static int O ();
static int P ();
static int Q ();
static int R ();
static int S ();
static int T ();
static int U ();
static int V ();
static int W ();
static int X ();
static int Y ();
static int Z ();
static int AA ();
static int BB ();
static int CC ();
*/
static int 
B (int stem_length, char *end)
{
  return (stem_length >= 3);
}

static int 
C (int stem_length, char *end)
{
  return (stem_length >= 4);
}

static int 
D (int stem_length, char *end)
{
  return (stem_length >= 5);
}

static int 
E (int stem_length, char *end)

     
{
  return (*end != 'e');
}

static int 
F (int stem_length, char *end)
     
     
{
  return ((stem_length >= 3) && (*end != 'e'));
}

static int 
G (int stem_length, char *end)
     
     
{
  return ((stem_length >= 3) && (*end == 'f'));
}

static int 
H (int stem_length, char *end)
     
     
{
  char p1, p2;
  p1 = *end;
  p2 = *(end - 1);
  return ((p1 == 't') || ((p1 == 'l') && (p2 == 'l')));
}

static int 
I (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((p1 != 'o') && (p1 != 'e'));
}

static int 
J (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((p1 != 'a') && (p1 != 'e'));
}

static int 
K (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((stem_length >= 3) &&
      ((p1 == 'l') || (p1 == 'i') ||
       ((p1 == 'e') && (*(end - 2) == 'u'))));
}

static int 
L (int stem_length, char *end)
     
     
{
  char p1, p2;
  p1 = *end;
  p2 = *(end - 1);
  if (p1 == 's')
    return (p2 == 'o');
  else
    return ((p1 != 'u') && (p1 != 'x'));
}

static int 
M (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((p1 != 'a') && (p1 != 'c') &&
      (p1 != 'e') && (p1 != 'm'));
}

static int 
N (int stem_length, char *end)
     
     
{
  if (stem_length >= 3)
    {
      if (*(end - 2) == 's')
    return (stem_length >= 4);
      else
    return 1;
    }
  else
    return 0;
}

static int 
O (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((p1 == 'l') || (p1 == 'i'));
}

static int 
P (int stem_length, char *end)
     
     
{
  return (*end != 'c');
}

static int 
Q (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;

  return ((stem_length >= 3) && (p1 != 'l') && (p1 != 'n'));
}

static int 
R (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;
  return ((p1 == 'n') || (p1 == 'r'));
}

static int 
S (int stem_length, char *end)
     
     
{
  char p1, p2;
  p1 = *end;
  p2 = *(end - 1);

  if (p1 == 't')
    return (p2 != 't');
  else
    return ((p1 == 'r') && (p2 == 'd'));
}

static int 
T (int stem_length, char *end)
     
     
{
  char p1, p2;
  p1 = *end;
  p2 = *(end - 1);

  if (p1 == 't')
    return (p2 != 'o');
  else
    return (p1 == 's');
}

static int 
U (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;

  return ((p1 == 'l') || (p1 == 'm') ||
      (p1 == 'n') || (p1 == 'r'));
}

static int 
V (int stem_length, char *end)
     
     
{
  return (*end == 'c');
}

static int 
W (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;

  return ((p1 != 's') && (p1 != 'u'));
}

static int 
X (int stem_length, char *end)
     
     
{
  char p1;
  p1 = *end;

  if (p1 == 'e')
    return ((stem_length >= 3) && (*(end - 2) == 'u'));
  else
    return ((p1 == 'l') || (p1 == 'i'));
}

static int 
Y (int stem_length, char *end)
     
     
{
  return ((*end == 'n') && (*(end - 1) == 'i'));
}

static int 
Z (int stem_length, char *end)
     
     
{
  return (*end != 'f');
}

static int 
AA (int stem_length, char *end)
     
     
{
  char p1, p2;
  p1 = *end;
  p2 = *(end - 1);

  if (p1 == 'h')
    return ((p2 == 'p') || (p2 == 't'));
  else if (p1 == 'r')
    return ((p2 == 'e') || (p2 == 'o'));
  else if (p1 == 's')
    return (p2 == 'e');
  else
    return ((p1 == 'd') || (p1 == 'f') ||
        (p1 == 'l') || (p1 == 't'));
}

static int 
BB (int stem_length, char *end)
     
     
{
  if (stem_length >= 3)
    {
      if (stem_length >= 4)
    return (strncmp (end - 3, "ryst", 4) != 0);
      else
    return (strncmp (end - 2, "met", 3) != 0);
    }
  else
    return 0;
}

static int 
CC (int stem_length, char *end)
     
     
{
  return (*end == 'l');
}

/**************************************************************************/

#define MIN_STEM_LENGTH 2
#define MAX_ENDING_SIZE 11
#define PREDEFINED_SIZE (MAX_ENDING_SIZE + MIN_STEM_LENGTH)
#define EOS     '\0'
/* [RPAP - Jan 97: Stem Index Change] */
#define MAX_STEM_LENGTH 255

static char *remove_ending (char *word, int w_len);
static void recode_stem (char *stem_end);

/* =========================================================================
 * Function: stem
 * Description: 
 *      Implemetation of the Lovins' stemming algorithm described in
 *      J.B. Lovins, "Development of a Stemming Algorithm",
 *      Mechanical Translation and Computational Linguistics, Vol 11,1968.
 * Input: a word string with the length in the first byte
 * Output: the stemmed word
 * ========================================================================= */

void
lovinstem (u_char * word)
{
  char *word_start;     /* points to first letter of word */
  char *stem_end;       /* points to last character of stem portion */
  int old_length;       /* length of word */
  int new_length;       /* length after stemming */

  /* [RPAP - Jan 97: Stem Index Change] */
  u_char copy[MAX_STEM_LENGTH + 2];
  memcpy (copy, word, *word + 1);

  old_length = *copy; /* [RPAP - Jan 97: Stem Index Change] */

  /* The word must be at least MIN_STEM_LENGTH characters long. */
  if (old_length <= MIN_STEM_LENGTH)
    return;

  word_start = (char *) copy + 1;   /* jump over length byte */ /* [RPAP - Jan 97: Stem Index Change] */
  copy[old_length + 1] = '\0';  /* null terminate string */ /* [RPAP - Jan 97: Stem Index Change] */

  stem_end = remove_ending (word_start, old_length);
  recode_stem (stem_end);

  /* fix up new length */
  /* have to use strlen because have no other way of telling length */
  new_length = strlen (word_start);
  *copy = new_length;  /* [RPAP - Jan 97: Stem Index Change] */

  /* [RPAP - Jan 97: Stem Index Change] */
  memcpy (word, copy, *copy + 1);

}               /*stem_complex */

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

 *   Data structure for recoding and the method for recoding the stem. 
 *
 * ========================================================================= */

typedef struct
{
  char *old_end;        /* old ending */
  char *new_end;        /* new ending */
  char old_offset;      /* length of the old ending - 1 */
  int (*cond) (char);           /* condition rule */
  char end_group;       /* signal the end of the group */
}
Recode_Rules;

/*  
   Stem terminations are divided into groups having the same last
   character
 */

static Recode_Rules Rules[] =
{
  "uad", "uas", 2, NULL, 0,
  "vad", "vas", 2, NULL, 0,
  "cid", "cis", 2, NULL, 0,
  "lid", "lis", 2, NULL, 0,
  "erid", "eris", 3, NULL, 0,
  "pand", "pans", 3, NULL, 0,
  "end", "ens", 2, s, 0,
  "end", "ens", 2, m, 0,
  "ond", "ons", 2, NULL, 0,
  "lud", "lus", 2, NULL, 0,
  "rud", "rus", 2, NULL, 1,

  "ul", "l", 1, aio, 1,

  "istr", "ister", 3, NULL, 0,
  "metr", "meter", 3, NULL, 0,
  "her", "hes", 2, pt, 1,

  "urs", "ur", 2, NULL, 1,

  "uct", "uc", 2, NULL, 0,
  "umpt", "um", 3, NULL, 0,
  "rpt", "rb", 2, NULL, 0,
  "mit", "mis", 2, NULL, 0,
  "ert", "ers", 2, NULL, 0,
  "et", "es", 1, n, 0,
  "yt", "ys", 1, NULL, 1,

  "iev", "ief", 2, NULL, 0,
  "olv", "olut", 2, NULL, 1,

  "bex", "bic", 2, NULL, 0,
  "dex", "dic", 2, NULL, 0,
  "pex", "pic", 2, NULL, 0,
  "tex", "tic", 2, NULL, 0,
  "ax", "ac", 1, NULL, 0,
  "ex", "ec", 1, NULL, 0,
  "ix", "ic", 1, NULL, 0,
  "lux", "luc", 2, NULL, 1,

  "yz", "ys", 1, NULL, 1
};

typedef struct last
  {
    char c;         /* The last character */
    struct last *left;      /* used in balanced   */
    struct last *right;     /* binary tree        */
    Recode_Rules *pt;       /* location of approriate group */
  }
Last_Char_Node;

/*------------
       s
     /   \
    l     x
   / \   / \
  d   r t   z
         \
          v
---------------*/
static Last_Char_Node pr[] =
{
  'd', NULL, NULL, Rules,
  'l', pr, pr + 2, Rules + 11,
  'r', NULL, NULL, Rules + 12,
  's', pr + 1, pr + 6, Rules + 15,
  't', NULL, pr + 5, Rules + 16,
  'v', NULL, NULL, Rules + 23,
  'x', pr + 4, pr + 7, Rules + 25,
  'z', NULL, NULL, Rules + 33,
};


/*****************************************************************************
 *
 *  Recode the stem according to the transformation rules.
 *
 *****************************************************************************/

static void
recode_stem (char *stem_end)
{
  static Last_Char_Node *root = pr + 3;
  Last_Char_Node *p_last = root;
  Recode_Rules *rule;       /* points to the transformation list */
  char *h,          /* points to start of possible ending */
    ch,             /* last character of the old ending  */
    ch2;

  h = stem_end - 1;
  ch = *stem_end;
  if (*h == ch)         /* Check for double consonant        */
    {
      if (strchr ("bdglmnprst", ch) != NULL)
    {
      *stem_end = EOS;
      stem_end = h;
    }
    }

  do                /* Check for the last character       */
    {
      ch2 = p_last->c;
      if (ch == ch2)
    break;
      else if (ch > ch2)
    p_last = p_last->right;
      else
    p_last = p_last->left;
    }
  while (p_last != NULL);

  if (p_last != NULL)       /* Check for the rest of suffix list */
    {
      rule = p_last->pt;
      for (;;)
    {
      h = stem_end - rule->old_offset;
      if (strcmp (h, rule->old_end) == 0)
        {
          if (!rule->cond || (*rule->cond) (*(h - 1)))
        {
          (void) strcpy (h, rule->new_end);
          break;
        }
        }
      if (rule->end_group)
        break;
      ++rule;
    }
    }
}

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

 *    List of endings and the method to remove the ending 
 *
 * ======================================================================*/

/************ Data structures for list of endings  ********************/

typedef struct
{
  char *ending;         /* old ending                */
  int (*cond) (int length, char *end);      /* conditional rule          */
  signed char left_offset;  /* used to find the siblings */
  signed char right_offset; /* in balanced binary tree   */
}
Ending_List;

/*
   The original list of endings is re-organised into groups having
   the same length and the same first character.
 */

static Ending_List List[] =
{
  "a", NULL, 0, 0,

  "ae", NULL, 0, 0,
  "al", BB, -1, 2,
  "ar", X, 0, 0,
  "as", B, -1, 0,

  "acy", NULL, 0, 1,
  "age", B, 0, 0,
  "aic", NULL, -2, 1,
  "als", BB, 0, 0,
  "ant", B, -2, 2,
  "ars", O, 0, 0,
  "ary", F, -1, 2,
  "ata", NULL, 0, 0,
  "ate", NULL, -1, 0,

  "able", NULL, 0, 1,
  "ably", NULL, 0, 0,
  "ages", B, -2, 2,
  "ally", B, 0, 0,
  "ance", B, -1, 1,
  "ancy", B, 0, 0,
  "ants", B, -4, 4,
  "aric", NULL, 0, 0,
  "arly", K, -1, 1,
  "ated", I, 0, 0,
  "ates", NULL, -2, 2,
  "atic", B, 0, 0,
  "ator", NULL, -1, 0,

  "acies", NULL, 0, 0,
  "acity", NULL, -1, 1,
  "aging", B, 0, 0,
  "aical", NULL, -2, 2,
  "alist", NULL, 0, 0,
  "alism", B, -1, 0,
  "ality", NULL, -3, 3,
  "alize", NULL, 0, 1,
  "allic", BB, 0, 0,
  "anced", B, -2, 2,
  "ances", B, 0, 0,
  "antic", C, -1, 0,
  "arial", NULL, -6, 6,
  "aries", NULL, 0, 1,
  "arily", NULL, 0, 0,
  "arity", B, -2, 2,
  "arize", NULL, 0, 0,
  "aroid", NULL, -1, 0,
  "ately", NULL, -3, 3,
  "ating", I, 0, 1,
  "ation", B, 0, 0,
  "ative", NULL, -2, 2,
  "ators", NULL, 0, 0,
  "atory", NULL, -1, 1,
  "ature", E, 0, 0,

  "aceous", NULL, 0, 1,
  "acious", B, 0, 0,
  "action", G, -2, 2,
  "alness", NULL, 0, 0,
  "ancial", NULL, -1, 1,
  "ancies", NULL, 0, 0,
  "ancing", B, -4, 4,
  "ariser", NULL, 0, 0,
  "arized", NULL, -1, 1,
  "arizer", NULL, 0, 0,
  "atable", NULL, -2, 2,
  "ations", B, 0, 0,
  "atives", NULL, -1, 0,

  "ability", NULL, 0, 1,
  "aically", NULL, 0, 0,
  "alistic", B, -2, 2,
  "alities", NULL, 0, 0,
  "ariness", E, -1, 0,
  "aristic", NULL, -3, 3,
  "arizing", NULL, 0, 1,
  "ateness", NULL, 0, 0,
  "atingly", NULL, -2, 2,
  "ational", B, 0, 0,
  "atively", NULL, -1, 1,
  "ativism", NULL, 0, 0,

  "ableness", NULL, 0, 1,
  "arizable", NULL, 0, 0,

  "allically", C, 0, 0,
  "antaneous", NULL, -1, 1,
  "antiality", NULL, 0, 0,
  "arisation", NULL, -2, 2,
  "arization", NULL, 0, 0,
  "ationally", B, -1, 1,
  "ativeness", NULL, 0, 0,

  "antialness", NULL, 0, 0,
  "arisations", NULL, -1, 1,
  "arizations", NULL, 0, 0,

  "alistically", B, 0, 1,
  "arizability", NULL, 0, 0,

  "e", NULL, 0, 0,

  "ed", E, 0, 0,
  "en", F, -1, 1,
  "es", E, 0, 0,

  "eal", Y, 0, 0,
  "ear", Y, -1, 1,
  "ely", E, 0, 0,
  "ene", E, -2, 2,
  "ent", C, 0, 0,
  "ery", E, -1, 1,
  "ese", NULL, 0, 0,

  "ealy", Y, 0, 1,
  "edly", E, 0, 0,
  "eful", NULL, -2, 1,
  "eity", NULL, 0, 0,
  "ence", NULL, -2, 2,
  "ency", NULL, 0, 0,
  "ened", E, -1, 2,
  "enly", E, 0, 0,
  "eous", NULL, -1, 0,

  "early", Y, 0, 1,
  "ehood", NULL, 0, 0,
  "eless", NULL, -2, 2,
  "elity", NULL, 0, 0,
  "ement", NULL, -1, 0,
  "enced", NULL, -3, 3,
  "ences", NULL, 0, 1,
  "eness", E, 0, 0,
  "ening", E, -2, 2,
  "ental", NULL, 0, 0,
  "ented", C, -1, 1,
  "ently", NULL, 0, 0,

  "eature", Z, 0, 0,
  "efully", NULL, -1, 1,
  "encies", NULL, 0, 0,
  "encing", NULL, -2, 2,
  "ential", NULL, 0, 0,
  "enting", C, -1, 1,
  "entist", NULL, 0, 1,
  "eously", NULL, 0, 0,

  "elihood", E, 0, 1,
  "encible", NULL, 0, 0,
  "entally", NULL, -2, 2,
  "entials", NULL, 0, 0,
  "entiate", NULL, -1, 1,
  "entness", NULL, 0, 0,

  "entation", NULL, 0, 0,
  "entially", NULL, -1, 1,
  "eousness", NULL, 0, 0,

  "eableness", E, 0, 1,
  "entations", NULL, 0, 0,
  "entiality", NULL, -2, 2,
  "entialize", NULL, 0, 0,
  "entiation", NULL, -1, 0,

  "entialness", NULL, 0, 0,

  "ful", NULL, 0, 0,

  "fully", NULL, 0, 0,

  "fulness", NULL, 0, 0,

  "hood", NULL, 0, 0,

  "i", NULL, 0, 0,

  "ia", NULL, 0, 0,
  "ic", NULL, -1, 1,
  "is", NULL, 0, 0,

  "ial", NULL, 0, 0,
  "ian", NULL, -1, 1,
  "ics", NULL, 0, 1,
  "ide", L, 0, 0,
  "ied", NULL, -3, 3,
  "ier", NULL, 0, 0,
  "ies", P, -1, 0,
  "ily", NULL, -1, 1,
  "ine", M, 0, 0,
  "ing", N, -5, 5,
  "ion", Q, 0, 0,
  "ish", C, -1, 1,
  "ism", B, 0, 1,
  "ist", NULL, 0, 0,
  "ite", AA, -3, 3,
  "ity", NULL, 0, 0,
  "ium", NULL, -1, 0,
  "ive", NULL, -1, 1,
  "ize", F, 0, 0,

  "ials", NULL, 0, 0,
  "ians", NULL, -1, 0,
  "ible", NULL, -1, 1,
  "ibly", NULL, 0, 0,
  "ical", NULL, -2, 2,
  "ides", L, 0, 0,
  "iers", NULL, -1, 1,
  "iful", NULL, 0, 0,
  "ines", M, -4, 4,
  "ings", N, 0, 0,
  "ions", B, -1, 1,
  "ious", NULL, 0, 0,
  "isms", B, -2, 2,
  "ists", NULL, 0, 0,
  "itic", H, -1, 1,
  "ized", F, 0, 1,
  "izer", F, 0, 0,

  "ially", NULL, 0, 0,
  "icant", NULL, -1, 1,
  "ician", NULL, 0, 0,
  "icide", NULL, -2, 2,
  "icism", NULL, 0, 0,
  "icist", NULL, -1, 0,
  "icity", NULL, -3, 3,
  "idine", I, 0, 1,
  "iedly", NULL, 0, 0,
  "ihood", NULL, -2, 2,
  "inate", NULL, 0, 0,
  "iness", NULL, -1, 0,
  "ingly", B, -6, 6,
  "inism", J, 0, 1,
  "inity", CC, 0, 0,
  "ional", NULL, -2, 2,
  "ioned", NULL, 0, 0,
  "ished", NULL, -1, 0,
  "istic", NULL, -3, 3,
  "ities", NULL, 0, 1,
  "itous", NULL, 0, 0,
  "ively", NULL, -2, 2,
  "ivity", NULL, 0, 0,
  "izers", F, -1, 1,
  "izing", F, 0, 0,

  "ialist", NULL, 0, 0,
  "iality", NULL, -1, 1,
  "ialize", NULL, 0, 0,
  "ically", NULL, -2, 2,
  "icance", NULL, 0, 0,
  "icians", NULL, -1, 1,
  "icists", NULL, 0, 0,
  "ifully", NULL, -4, 4,
  "ionals", NULL, 0, 0,
  "ionate", D, -1, 1,
  "ioning", NULL, 0, 0,
  "ionist", NULL, -2, 2,
  "iously", NULL, 0, 0,
  "istics", NULL, -1, 1,
  "izable", E, 0, 0,

  "ibility", NULL, 0, 0,
  "icalism", NULL, -1, 1,
  "icalist", NULL, 0, 1,
  "icality", NULL, 0, 0,
  "icalize", NULL, -3, 3,
  "ication", G, 0, 0,
  "icianry", NULL, -1, 0,
  "ination", NULL, -1, 1,
  "ingness", NULL, 0, 0,
  "ionally", NULL, -5, 5,
  "isation", NULL, 0, 0,
  "ishness", NULL, -1, 1,
  "istical", NULL, 0, 1,
  "iteness", NULL, 0, 0,
  "iveness", NULL, -3, 3,
  "ivistic", NULL, 0, 0,
  "ivities", NULL, -1, 0,
  "ization", F, -1, 1,
  "izement", NULL, 0, 0,

  "ibleness", NULL, 0, 0,
  "icalness", NULL, -1, 1,
  "ionalism", NULL, 0, 0,
  "ionality", NULL, -2, 2,
  "ionalize", NULL, 0, 0,
  "iousness", NULL, -1, 1,
  "izations", NULL, 0, 0,

  "ionalness", NULL, 0, 1,
  "istically", NULL, 0, 0,
  "itousness", NULL, -2, 2,
  "izability", NULL, 0, 0,
  "izational", NULL, -1, 0,

  "izationally", B, 0, 0,

  "ly", B, 0, 0,

  "less", NULL, 0, 1,
  "lily", NULL, 0, 0,

  "lessly", NULL, 0, 0,

  "lessness", NULL, 0, 0,

  "ness", NULL, 0, 0,

  "nesses", NULL, 0, 0,

  "o", NULL, 0, 0,

  "on", S, 0, 1,
  "or", T, 0, 0,

  "oid", NULL, 0, 0,
  "one", R, -1, 1,
  "ous", NULL, 0, 0,

  "ogen", NULL, 0, 0,

  "oidal", NULL, 0, 0,
  "oides", NULL, -1, 2,
  "otide", NULL, 0, 0,
  "ously", NULL, -1, 0,

  "oidism", NULL, 0, 0,

  "oidally", NULL, 0, 1,
  "ousness", NULL, 0, 0,

  "s", W, 0, 0,

  "s'", NULL, 0, 0,

  "um", U, 0, 1,
  "us", V, 0, 0,

  "ward", NULL, 0, 1,
  "wise", NULL, 0, 0,

  "y", B, 0, 0,

  "yl", R, 0, 0,

  "ying", B, 0, 1,
  "yish", NULL, 0, 0,

  "'s", NULL, 0, 0,
};

typedef struct node
  {
    char c;         /* First character  */
    struct node *left;      /* used in balanced */
    struct node *right;     /* binary tree      */
    Ending_List *ptr[11];   /* the approriate location */
  }
First_Char_Node;

static First_Char_Node First[] =
{
  '\'', NULL, NULL,
  {NULL,
   List + 293, NULL, NULL, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},

  'a', First, NULL,
  {List,
   List + 2, List + 9, List + 20, List + 39, List + 58,
   List + 70, List + 77, List + 82, List + 87, List + 89},

  'e', First + 1, First + 4,
  {List + 91,
   List + 93, List + 98, List + 106, List + 116, List + 126,
   List + 133, List + 138, List + 142, List + 145, NULL},

  'f', NULL, NULL,
  {NULL,
   NULL, List + 146, NULL, List + 147, NULL,
   List + 148, NULL, NULL, NULL, NULL},

  'h', First + 3, First + 5,
  {NULL,
   NULL, NULL, List + 149, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},

  'i', NULL, NULL,
  {List + 150,
   List + 152, List + 163, List + 181, List + 202, List + 222,
   List + 239, List + 252, List + 258, NULL, List + 261},

  'l', First + 2, First + 10,
  {NULL,
   List + 262, NULL, List + 263, NULL, List + 265,
   NULL, List + 266, NULL, NULL, NULL},

  'n', NULL, NULL,
  {NULL,
   NULL, NULL, List + 267, NULL, List + 268,
   NULL, NULL, NULL, NULL, NULL},

  'o', First + 7, First + 9,
  {List + 269,
   List + 270, List + 273, List + 275, List + 277, List + 280,
   List + 281, NULL, NULL, NULL, NULL},

  's', NULL, NULL,
  {List + 283,
   List + 284, NULL, NULL, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},

  'u', First + 8, First + 12,
  {NULL,
   List + 285, NULL, NULL, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},

  'w', NULL, NULL,
  {NULL,
   NULL, NULL, List + 287, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},

  'y', First + 11, NULL,
  {List + 289,
   List + 290, NULL, List + 291, NULL, NULL,
   NULL, NULL, NULL, NULL, NULL},
};


/******************************************************************************
 *
 *  Look for a match in the suffix list.
 *  Return the pointer to the end of the new stem if there is a match.
 *  Otherwise, return the pointer to the end of the original word.
 *
 *  Method:
 *      * Search for the first character in the suffix list.
 *      * If there is match, search for the rest of the suffix list.
 *         
 *****************************************************************************/

static char *
remove_ending (char *word, int w_len)
{
  static First_Char_Node *root = First + 6; /* the root of binary tree is 'l' */
  First_Char_Node *p_first; /* Points to the first character      */
  /* of the possible suffix             */
  Ending_List *p_list;      /* Points to the List of Endings      */
  char *suffix_start,       /* Points to start of possible suffix */
   *stem_end,           /* Points to the end of possible stem */
    ch1, ch2;
  int s_len,            /* Possible stem length               */
    e_offset,           /* Offset from the end to the start of */
  /* possible suffix                    */
    not_done = 1, cmp;

  stem_end = word + w_len - 1;
  if (w_len > PREDEFINED_SIZE)  /* find the start of suffix   */
    suffix_start = word + w_len - MAX_ENDING_SIZE;
  else
    suffix_start = word + MIN_STEM_LENGTH;

  ch1 = *suffix_start;
  do                /* Search for the first character     */
    {
      p_first = root;
      do
    {
      ch2 = p_first->c;
      if (ch1 == ch2)
        break;
      else if (ch1 > ch2)
        p_first = p_first->right;
      else
        p_first = p_first->left;
    }
      while (p_first != NULL);

      if (p_first != NULL)  /* Search for the rest */
    {
      e_offset = stem_end - suffix_start;
      if ((p_list = p_first->ptr[e_offset]) != NULL)
        {
          for (;;)      /* no need to compare the first char  */
        {
          cmp = strcmp (suffix_start + 1, p_list->ending + 1);
          if (cmp > 0)
            if (p_list->right_offset)
              p_list += p_list->right_offset;
            else
              break;
          else if (cmp < 0)
            if (p_list->left_offset)
              p_list += p_list->left_offset;
            else
              break;
          else
            {
              s_len = suffix_start - word;
              if (!p_list->cond ||
              (*p_list->cond) (s_len, suffix_start - 1))
            {
              *suffix_start = EOS;
              stem_end = suffix_start - 1;
              not_done = 0;
            }
              break;
            }
        }
        }
    }
      ++suffix_start;
    }
  while (not_done && ((ch1 = *suffix_start) != EOS));
  return (stem_end);
}