This directory contains the files for a comprehensive spelling checker which
can correct files containing text and embedded commands for a number of
different word processors. All necessary files are provided, including
dictionaries. This file contains a full description of the program and
installation instructions.

Files:
*****

AAAREADME.TXT	- this file
BUILD.COM	- installation procedure
COMMONWRD.DAT	- sequential file of most common words
LEXIC08.DAT	- main dictionary for words of <8  letters - indexed file
LEXIC16.DAT	- main dictionary for words of <16 letters - indexed file
LEXIC32.DAT	- main dictionary for words of <32 letters - indexed file
NEWGOOD.TXT	- new word file - indexed
SPELL.EXE	- SPELL program compiled under VMS v3.7
SPELL.HLP	- HELP library module for SPELL
SPELL.OBJ	- object file of SPELL
SPELL.PAS	- VMS Pascal source text of SPELL

Total of 12 files, 4263/4343 blocks.

Functionality:
*************

SPELL reads standard VMS text files - such as those created by EDT, and will
recognize the syntax for embedded commands in RUNOFF, TeX and SCRIBE word
processor input files, thus preventing many spurious "errors" from being
registered. The program makes extensive use of VT100 screen formatting
abilities, and as supplied, can only be used on VT100-compatible terminals.
The entire program is written in Pascal, and should be readily modifiable
for other terminals, or if additional functionality is required.

The checker includes guessing algorithms which may optionally be invoked if a
wrongly spelled word is encountered, a facility for creating personal
dictionaries, and provision for the automatic submission of words not found in
the dictionaries to a maintenance person so that the main dictionaries may be
periodically updated. The checker creates a corrected file as output, and
optionally also produces a log file showing all changes made. 

Installation:
************

There is an installation procedure in this submission which should make
installation relatively painless. Unlike SPELL, it is not extensively
error-trapped, so be careful when using it - hopefully, you will only need to
use it once! To use this procedure, copy all the files in this directory
to an empty directory on your system (not necessarily on your system disk),
set your default to that directory, and invoke the procedure with an @BUILD
command.

Re-installation:
***************

If you ever need to re-install SPELL (for instance because of an operating
system upgrade), do NOT use the BUILD procedure provided without looking
through it carefully, and commenting out any sections which deal with
updating the dictionary files. Unless you do this, you will create new
copies of the dictionaries, and will lose all the site-specific words
that you have added since you first installed the program! This caveat
also applies to sites running the Temple PROOFREAD program (see "Origins"
below).

Internals: 
*********

Spell operates by constructing an ordered tree of all new words as they are
encountered in a document. Any word in the document is first checked against
this tree, and if it is already in the tree, it is copied intact. If the word
under test is not in the tree, and it is not one of a small number of very
common words which are held in an array in memory, one of the three indexed
files which constitute the main dictionary is searched, depending on the length
of the word. If the word is still not found, an optional personal dictionary in
the user's default directory is searched, and failing this, the list of words
which users have marked as "right" but which have not yet been checked by the
dictionary-person. If the word appears in this last list, the user is warned
that it may not be correct, and given the opportunity to reject it.

If a word is not found in any of these searches, the user is presented with
two lines of context, with the word highlighted, and given a number of
choices:

  C - you want to check alternative spellings in the  dictionary.  I will
      prompt you for words to check until you find one that is correct or
      you decide to give up checking.
  
  G - you want me to offer you some suggestions about possible spellings.
  
  I - you want to ignore the word - for instance an uncommon name.
  
  P - you want me to add the word to your Personal Dictionary.
  
  Q - you want to quit SPELL without checking any more of the file.
  
  R - the word is right and should be added to the main dictionary.
  
  W - you  agree that the spelling of  the word is wrong.  I will ask you
      for the correct spelling or give you the option of asking me to guess.

If the result of a "wrong", "guess" or "check" operation is to change the
word being questioned by spell, the program will attempt to automatically
match upper and lower case to the case of the word being replaced. If this
turns out to be impossible, the user is prompted to supply the case.

Wrong words are linked into a double tree together with replacements. If
they occur again in the file, the user is prompted to decide whether
they wish to make the same change again.

Right words are added to an indexed file in the system dictionary
directory, which is automatically accessed when the dictionary person
invokes the SPELL/UPDATE command.

SPELL writes a workfile called SPELLWRK.TMP, and if the session terminates
successfully, and any changes have been made to the file being checked,
then the workfile is renamed to the next higher version of the file being
checked.

The "front-end" of SPELL is a small DFA lexical scanner. This can easily be
modified to include support for other word processors. All pre-processing of
text is handled by this scanner, including recognition of words, constructs
such as apostrophes and quotations, and recognition of embedded word-processor
commands. Any word or character which is not recognized as a possible
"legitimate" word is simply copied to the output file, and the remaining
candidates for testing are made into a simple linked-list which contains some
attributes of the words as well as their text, and allows simple manipulation
of the text if errors are encountered. 

As supplied, SPELL operates with Control-Y disabled. This is not strictly
necessary, given that an intermediate file is used for output, but in
practice, we have found this to be a good idea in avoiding problems with
accidental exits from the program. If you don't like this idea, taking
out the call to the disabling routine is trivial.

Almost all terminal I/O is handled using direct calls to QIO rather
than Pascal I/O (sigh) since the latter is generally unsuitable for the
types of screen and keyboard operations involved.

Origins:
*******

The SPELL program is based on a program called PROOFREAD which was written
by Matthew Temple of Smith College, Northampton MA. His ideas and permission
to use them are gratefully acknowledged.

Submitted by:
********* **

Mark Resmer, Vassar College, Poughkeepsie NY 12601 (914) 452-7000 Ext 2437

resmer@vassar on CCnet and BITnet