readme.act

ears-0.32, linux下有用的语音信号处理工具包

From [e-mail] May 28 09:09:47 1995
Date: Sat, 27 May 1995 07:28:14 -0400 (EDT)
From: Dan Chaney <[e-mail]>
To: [e-mail]
Subject: README for act.c


[ Here is a program that parses the output of 'listen' and provides a
  VERY rudimentary C-program to build and execute commands with 'listen'.
  This code is fairly atrocious, but if I don't send it on to people on
  the [e-mail] soon, it will be of even less use.  -dan ]

README file for act.c		5/95			[e-mail]


  This is the README file, the sections are:
	- What this is (and why)
	- What it does, briefly
	- Definitions (Because sometimes, I make no sense...)
	- What to do to build the program: (This is what you're looking for)
	- List of words to say (and the tokens they return)


- What this is (and why)
This program is designed to run with Ralf W. Stephan's <[e-mail]>
'listen' program from his ears package.  This program, 'act', parses the
output of 'listen' and builds a sentence based on the output.  That
sentence may be executed, corrected or dumped entirely.  This code is so
young (and so bad :) I suspect it will be something to tinker with and
then say 'I could do that better if....'  Please update the rest of us
when you do :-)

Why?  Because I don't know c++ well and wanted a C program to tinker
with in order to get the voice recognition to execution shell commands.

- What it does (briefly)
I found parsing is a much easier thing if you have the words printed
by listen (I call them tokens) encased in brackets.  Single character
tokens such as letters, numbers and punctuation are more easily handled 
if they are marked as such, (ie [letter-0], [letter-1], [letter-~], etc.)

As a word is spoken and recognized, the []-delimited token is printed by 
'listen' and read by 'act'.  'Act' checks to see if the token has special
meaning (see 'command tokens') and, if not, adds it into the sentence.
Unless the token is marked as a letter (ie, letter-d], a space is added
into the sentence between tokens.  (A space may be specically added, using
the [space] token.  Spaces and dashes are important to the Unix command
line, so they are handled slightly differently.)

So, given a set of []-delimited words, 'act' will build a string (even
allowing you to backspace verbally by saying 'wrong' or 'backspace'
or even 'clear'.

- Definitions (Because sometimes, I make no sense...)
	Token - the string 'listen' prints out in response to a word.
		'act' will ignore anything not in []'s
	Sentence - the list of tokens printed from 'listen' to 'act'
	Commands - tokens that 'act' performs on
	
- What to do to build the program: (This is what you've been looking for)

	1. Set your default.words file to output the words within brackets.
	   (Edit default.words, put the second word in brackets like [this])

	2. Check the list of tokens below.  'Act' knows to watch for them
	   and some (marked below) are special to 'act'.  Otherwise, the
	   word is just added into the sentence (and may be executed).

	3. Compile the program up (gcc act.c -o act) and run it (listen | act)


- List of words to say (and the tokens they return)

What you say:		listen's output           Comments
-------------		---------------	    -----------------------
edit 			[vi]		    Note, two words can have the same
vi 			[vi]		    token if their meaning is the same
exit 			[quit]		 
quit 			[quit]		 
list 			[less]
more 			[less]
less 			[less]
no 			[no]		 
yes 			[yes]		 
return 			[return]
logout 			[logout]
cdplayer 		[cdplayer]	Bad choice, play a CD, make more noise
play 			[play]		My alias for cdplayer
info 			[info]		Used with cdplayer
stop 			[stop]		Used with 'cdplayer'
volume 			[vol]		Adjusts speaker volume
phone 			[phone]		Shell script to show phone's status
uptime 			[uptime]

--------------- These are tokens with special meaning to 'act' ------------
execute 		[execute]	Send current sentence to system() call
doit 			[execute]
wrong 			[wrong]		Remove last token from sentence
space 			[space]		Add a space into the sentence
dash 			[-]		Add a dash into the sentence
backspace 		[backspace]	Remove one space from sentence
clear 			[clear]		Clear current sentence (start over)
show 			[show]		Show current sentence
abort 			[abort]		Exit 'listen' completely
restart 		[restart]	Tries to restart 'listen | act' command
mute 			[mute]		Toggle whether words are added to 
                                        the sentence.
computer 		[computer]	Check to see if program is running


--------------- These are tokens that 'act' converts to commands -----------
call 			[call]		Convert 'call' into 'cu -l /dev/ttyS2'
xterm 			[xterm]		Convert 'xterm' into 'xterm &'
Mosaic 			[Mosaic]	Convert 'call' into 'Mosaic &'

--------------- These are single-character tokens for spelling and numbers
slash 			[letter-/]
dot 			[letter-.]
tilde 			[letter-~]
Zero 			[letter-0]		         
One 			[letter-1]			 
Two 			[letter-2]		         
Three 			[letter-3]
Four 			[letter-4]
Five 			[letter-5]
Six 			[letter-6]
Seven 			[letter-7]
Eight 			[letter-8]
Nine 			[letter-9]
alfa 			[letter-a]
bravo 			[letter-b]
charlie 		[letter-c]
delta 			[letter-d]
echo 			[letter-e]	bad choice since its a Unix command
foxtrot 		[letter-f]
golf 			[letter-g]
hotel 			[letter-h]
india 			[letter-i]
juliet 			[letter-j]
kilo 			[letter-k]
lima 			[letter-l]
mike 			[letter-m]
november 		[letter-n]
oscar 			[letter-o]
papa 			[letter-p]
quebec 			[letter-q]
romeo 			[letter-r]
sierra 			[letter-s]
tango 			[letter-t]
uniform 		[letter-u]
victor 			[letter-v]
whiskey 		[letter-w]
x-ray 			[letter-x]
yankee 			[letter-y]
zulu 			[letter-z]
comma 			[letter-,]

Comments, suggestions and code updates welcome, [e-mail]

-- 
-- Daniel Chaney        From home, work or play...       Fax: [phone]   --
-- 715 Patterson Office Tower, Lexington, KY 40526-0027       606-257-2150   --
-- {uunet and the like}!ukma!chaney  [e-mail]  [e-mail]      --
--     "I'll have time enough for sleep when I'm dead and in the ground"     --