File GMLRules.txt  V. 2.0   11/12/96  NIST/WMF
  
  1. General.
  
     The rules that program "rfilter1" uses are simple string-rewriting
  rules of the form
  
      A => B
  or
      A => B / C __ D
  
    where A, B, C, and D are character strings.
  
     Each of these strings may contain internal spaces.  Strings may be
  bounded with either square brackets or single quotation marks, in
  order to allow beginning and ending spaces.
  
     The main purpose of using rules like this is to provide
  documentation of a string-rewriting process that is easily understood.
  
     Examples:
  
       CANCELLED => CANCELED ;; per AHD
       [  ] => [ ] ;; reduce 2 spaces to 1
       Falkner => Faulkner / [William ] __ 
       JETLINER => JET LINER
       VIDEOTAPE => VIDEO TAPE / [ ] __ [ ]
  
   2. Application Algorithm.
  
     In rewriting the input string into the output string, a character
  cursor is moved from the start to the end of the input string.  At
  each position, the list of rules in the specified rule file is
  searched from top to bottom.  The first rule (if any) that is found to
  match the input string with the first character of its input field (A)
  lined up at the cursor position is applied by concatenating its output
  field (B) onto the output string and advancing the cursor by the
  number of characters in the rule's input field (A).  If no rule
  matches, then the input character pointed to by the cursor is either
  passed on to the output string or ignored, depending on a switch
  setting, and the cursor position is advanced by one.
  
     An indexing scheme is used for speed, but the logical effect is
  still as if a linear search were done on the rules in their order as
  they are in the rules file.
  
    3. Comments.
  
     The token denoting comments is taken to be the first token in the
  first line of the file of rules.  Anything on a line following this
  comment flag is ignored by the program.
  
     4. Header Information.
  
     At the beginning of a file of rules, certain information may be
  given in keyword/value format in "auxiliary" lines that begin with
  "*".  The value must be the only token in the line bounded by quotes,
  either single or double; the keyword itself just must be some token on
  the line.
  
     Here are the current keywords and values:
  
    KEYWORD     VALUE
     NAME       The name of the rules, for documentation.
     DESC       A description of the rules, for documentation.
    FORMAT      NIST1 : context-free rules only (the default)
                NIST2 : context-sensitive rules
  MAX_NRULES      <N> : the number of rules to allocate space for.
  COPY_NO_HIT     "T" / "YES"/ "TRUE" : if no rule hits, copy input
                  "F" / "NO" / "FALSE": if no rule hits, skip input
  CASE_SENSITIVE  "T" / "YES" / "TRUE" : case-sensitive matching.
                  "F" / "NO" / "FALSE" : case-insensitive matching.
  
    Both the keyword and its value (when alphabetic) may be upper- or
  lower-case.
  
    Examples:
  
     * NAME = "spcor1.rls"
     * DESC : "Spelling Correction Rules #1"
     * FORMAT = 'NIST2'
     * MAX_NRULES = '200'
     * COPY_NO_HIT = 'T'
     * CASE_SENSITIVE = 'F'
  
  
  History: 
    V2.0 - part of the tranfilt distribution as rules.doc
    V2.1 - Moved to SCTK