COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002		WRITEUP ON SAYTST.SAI/DMP[1,REM]/[1,3] AND SAY1.DAT[1,REM]
C00012 00003		SOME TIPS ON USING THE PHONEMES WHEN CREATING A DICTIONARY
C00017 00004	∂07-OCT-74  1402		H,RF
C00019 ENDMK
C⊗;
	WRITEUP ON SAYTST.SAI/DMP[1,REM]/[1,3] AND SAY1.DAT[1,REM]

Start program by typing "RU SAYTST[1,REM]" for a fresh copy with
no dictionary loaded yet, or "R SAYTST" for a version with a standard
dictionary already loaded to save you time.

SAYTST will gronk the voder using whatever text string you type in
unless it begins with a special character (@ ! or ←), and append to
the output file SAYTST.DAT on your disk area as mentionned below.
See SAY.DOC[1,BO] for description of phoneme abbreviations available.
This program can be run while other users are running the voder --
It only INIT's the voder when actually sending data to it.
You can hear other voder users by doing <esc>10U on your console,
or hear only your own voder usage by doing <esc>U.

		-- NORMAL LINES OF INPUT TEXT --

The first character of each line you type for normal input determines what is
done with that line.  If it is not one of the special characters listed
below, the entire line is sent to the voder as-is.

/ -- Remainder of the line is a command to modify internal parameters,
most of which are bit-wise masks of related switches.  /? to see a list
of them with an explanation of each bit, and their current values.

? -- Remainder of line is ignored and this list of first-characters is typed.

@ -- Remainder of line is a file name in usual format <f>.<e>[pj,pg].
The file is expected to be a dictionary in the format
<english> <tab> <transliteration or macro> <tab> <comments>.
The in-core dictionary is updated by the contents of the file.
@SAY1.DAT[1,REM] will get REM's dictionary, in which in the case of two
different transliterations of a word (i.e. different but neither one
definitely better than the other), one is in the dictionary with the
string "'X" appended to the English.  If you think the 'X version
is better than the normal version, send REM a note with specifics,
otherwise the 'X versions might get eventually purged.

The program will detect adjacent definitions of the same word and write a
warning in SAYTST.DAT, a feature most useful if the dictionary file is in
alphabetical order.  If /D1 or /D2 is in effect (/D1 is default)
it will recite both versions as it recognizes the trouble.  If /D2 is
in effect you may manually select which one goes into core.

! -- Remainder of line is transliterated using the in-core dictionary.
If the transliteration of a word begins with ! it is a macro which is
transliterated again.  The program warns you if this recursive macro
expansion generates anything gigantic, and if you think it is in an
infinite loop, hitting <ALTMODE> will make it quit prematurely.

← -- Remainder of line is a file name.  The file is recited as if
every line of it were typed in with ! in front of it.  In addition,
if /R2 is in effect it breaks on punctuation rather than <CRLF>.
It will ask you for page number and starting line number or line
range in format <start>!<count>, but if you give
a large count you can abort by hitting <ALTMODE> almost anytime.

	-- ACTIVATION CHARACTER AT END OF INPUT LINE --

The <CR> key is the normal way to terminate a line of input.  If your
activation character is any of those listed below, something special
happend in addition to what is caused by the text of the line.
(If a strange activation character is hit otherwhere than the end of
the line, characters after the activation character are usually lost
-- I don't feel like taking great pains to recover those characters.)

<CTRL><META><PERIOD> -- the output file SAYTST.DAT
is updated fully before the input line is acted-upon.

<CTRL><META>Q -- (don't use this unless you know what you are doing)
This special command causes the program to exit after modifying its
starting address to be a continuation address in effect.  The normal
use of this is to save the current program, with dictionary already
loaded in, on the disk, to save time later when restarted.  To use this
feature, first load in all the dictionaries you normally use, then
use this command, then save the core image on a file (use ssave if you
have a version with an upper segment, save otherwise).  Later, starting
this core image (by the RU command or the GET and S commands) will actually
continue rather than start.

<CTRL><META><QUESTIONMARK> -- these special activation characters are listed.

   -- RECOVERING FROM LOSING YOUR LINE EDIT, AND UPDATING DICTIONARY --

After you type in anything, the program will attempt to restore the
most recent nonblank line you have typed, to your line editor
which will normally be the thing you have just finished typing.
Thus that you can edit it again without hitting <CTRL><CR> yourself. In fact
it is better than <CTRL><CR> because it goes back to the last nonblank
line of input it has seen and restores that, even if you have typed some
funny control characters in the meantime.  This means that if you want
to roll back to the previous line of input you just type <CLEAR><CR>.
This roll-back also writes the restored line onto the disk file SAYTST.DAT
on the assumption that if you rolled back to it once, you might want it
permanently saved on the disk too (see next paragraph for more on this
process).  The last ten nonblank lines you have
typed are kept on a push-down list, and the roll-back causes one line
to be popped-up, so if you want to roll back further all you have to do
is type <CLEAR><CR> again, up to ten times.

When you find a pronunciation of a word that you like, you may save it
on the disk by using the above <CLEAR><CR> feature, i.e. by rolling back
to it.  You may also save macros by the same method.  The program will
always ask you for the English that corresponds to the voder or macro,
and you have three options.  If you hit altmode, nothing is saved but
the line-editor is rolled-back.  If you hit <CR> with a blank line,
it writes the line-image to disk as is, and ditto line editor.  If you
type something followed by <CR>, it appends <something><tab> in front
of the string before writing to disk, and also enters the transliteration
or macro into the in-core dictionary, and ditto line editor.

	SOME TIPS ON USING THE PHONEMES WHEN CREATING A DICTIONARY
	(THIS PAGE IS NEW AND NOT GUARANTEED TO BE ACCURATE)

First of all, everyone should know about dipthongs, both of the vowel and
consonant type.  For example, the "A" in "TAKE" is really a short "E" as
in "BETH" followed by a long "E" as in "NEED".  The "J" in "JOSEPH" is
really a "D" as in "DAVIS" followed by a "ZH" as in "AZURE".  Most of the
dipthongs can be voded by sounding the two components in sequence.  Some
can also be voded by special dipthong pseudo-phonemes that BO's voder
driver somehow knows about.  If you sound out each troublesome word
very very slowly, you will probably be able to hear the two sounds in
a dipthong, and be able to generate voder phonemes accordingly.

Some sounds are easy to make on the voder and almost always come out
just the way they ought to, providing you have memorized the table in
SAY.DOC[1,BO], especially "X", "x", "U" and "u".
	F as in FUNNY, LAUGH, LAUGHTER -- use F
	H as in HELLO -- use H
	K as in KILL, HACKER -- use K
	L as in LOVE, ALSO, HALL -- use L
	M as in MONEY, HUMOR, DRUM -- use M or MM
	N as in NIL, SUNDAY, SUN -- use N or NN
	NG as in SINGER, SING -- use n or nG
	P as in PETER, UPPER, UP -- use P
	S as in SAY, PASSER, PASS -- use S (at end of word, try SZ) (try also Z)
	Z as in ZOO, SIZZLE, AS -- use Z (at end of word, try ZS) (try also S)
	SH as in SHOE, FISHER, ASH -- use s
	ZH as in AZURE -- use z
	T as in TABLE, BETTER -- use T or d
	TH as in THINK, BATH -- use X
	TH as in THEN, OTHER, BATHE -- use x (see below also)
	W as in WALL, AWOL -- use W
	Y as in YES, AYER -- use Y

	a	as in FAT
	e	as in BET
	i	as in MISS
	u	as in TUNNEY

	E	as in NEED	(most long vowels need their length carefuly
	O	as in SMOKE	  trimmed, such as "E1E2" instead of just "E")
	U	as in TUNE

Some sounds are a little difficult and need some help:
	T as in BET -- use Th
	D as in WED -- use Dh (see also below)
	K as in HACK -- use Kh

Some sounds are not obvious at all, or have several almost-equivalent
ways to do them with enough subtle differences to make the difference
between a passible and a good sound:
	OO as in BOOK -- compare R3R3R3, R3R3, o1
	A = eE as in WAIT -- compare e2E1, e2E2, A, A1, A2, ∃, e3A2
	ER as in VODER -- compare R, R3R, u3R, e2R
	R as in RUN -- compare R1, R2 -- don't use R
	AA as in MAAS, BOSS, COST, FATHER -- compare ∀, α, o
	J = DZH as in JUST -- compare Dz, Tz, J
	CH = TSH as in CHAIR -- compare Ts, C
	E as in SEATED -- compare E, i3y1, i2y1

Some sounds don't come out right no matter what we do.  It seems the voder
just isn't capable of making the correct sound.  If anyone comes up with
an acceptable way of pronouncing any of these, send REM a note.
	D as in AIMED, ALARMED, BED, CALLED --
	B as in BACK, BAD, BE, BRING, LAB --
	G as in BAGS --
	V as in HAVE --
	TH as in ALTHOUGH --

∂07-OCT-74  1402		H,RF
 RANDOM THOUGHTS CONCERNING EFFECTIVE TRANSCRIPTIONS
 FROM ENGLISH TO VODISH:
 
 1)	Never use R2 or L2.  The sounds are quite weak.
 
 2)	To get consonental R, L, M, N use R1, L1, M1, N1.
 
 3)	An unaccented vowel should always have length 2 or 3.
 
 4)	A vowel followed by R, L, M, N should use a very short
 vowel (if any) and let the liquid consonant fill up the rest.
 
 5)	An accented vowel should have length 0 or, better 1.
 
 6)	Most words sound best when run-on, without any space between
 words.
 
 7)	Some consonants, particularly stops like d, D, G, T, K
 have a different sound when followed by a space or some letter.
 When making dictionaries, do not be fooled by this; do not put
 the space in the dictionary.
 
 8)	The various lengths of vowels have different sounds as
 well as lengths.  Most noticible is the difference between
 o and o1.
 
 9)	some useful correspondences:
 	OO as in BOOK:		o1
 	J  as in JERRY:		Dz or Tz
 	EA as in SEATED:	i3y1 or i2y1  (dont use E in general)
 	A  as in TAKE:		e3A2