;.l1,3,63,66,1,0,10,75,192,2,15,20,25,127,10,0,
;.l2,15,75,192,2,20,25,127,15,0,
;.l3,20,75,192,2,25,127,20,0,
;.l4,25,75,192,2,127,25,0,
;.l5,30,75,192,2,127,30,0,
;.l6,12,90,192,2,18,24,127,12,1,
;.l7,17,124,192,2,34,127,17,2,
TO:    Dist., those working on collecting teachings of R. Shlomo
FROM:  Steve Amdur, Kibbutz Haon, Jordan Valley, 15170 Israel.
       Telephone:  972-6-757572; FAX:  972-6-757554
DATE:  12/29/94
RE:    Transcribing machine/ Flat-bed OCR Scan/ Voice Recognition
REF:   PC Magazine December 20, 1994
	  See review of HP Scanjet 3p, p38-39
CC:    Eliahu Gal-Or, Moshav Meor Modi'in  
------------------------------------------------------------------

A)  Safeguarding tapes:  I suggest that all available tapes beinventoried and backed up with 2 copies , that the original besaved as archival, that 1 lst-generation backup also be saved asarchival (for fear of tape aging), and that transcription only bemade from non-original tapes.  
	Some tapes could, with an edit (as videotapes are edittedbefore being distributed), be suitable for distribution andpossible sale.  
	In some cases, good tape-playing/copying equipment will benecessary before the entire tape can be described; R. Shlomo'svoice often drops to a nearly inaudible level (especially relativeto the background tape "white" noise, which increases withsubsequent tape generations) when he makes his most importantpoints. 
	I assume there are programs & systems for pulling out voicefrom background white noise.
	Ben-Zion Solomon (Moshav Mevo Modi'in, 73122) has expertisein these areas.
	Again, it's important that original tapes be adequatelybacked up and safely stored until they can be transcribed;transcribing originals on poor equipment could cause loss ofmaterial.
                                    
1)  TRANSCRIBING MACHINE
Again:  Anyone transcribing a large number of R. Shlomo tapes --pre-eminently, the Witts -- needs a foot-pedal-operatedtranscribing machine, with adaptor for standard audio cassette(rather than the mini-cassette used in office dicatating machine),with adjustable automatic reverse.  I'd guess that using such amachine, rather than the portable cassette recorders we now use,would triple productivity.  I've not seen them in Israel, thoughthey must be somewhere, new or used.   They were standard (thousually without the cassette adaptor) in USA offices 10 years ago;I don't know if they still are.

2) OCR SCANNING
There are a large number of non-input typed transcripts, primarily(as far as I know) in the Witt collection.  These should be input. In addition, the R. Shlomo material in HBG and CNS (Connections)should be input.  The options are re-typing and SCANing.  SCANingis less work, and probably cheaper if one figures on payingtypists the standard per-page rate.  Some commerical firms doScanning on a per-page or per-hour (customer operates machine)basis; I have a NIS 500 credit which anyone is welcome to use forthis purpose (on condition that, without prejudicing copy-rights, scanned input be shared amongst those working on R. Shlomomaterial) with Kivun haMa'arav, 43 Emek Refaim, Jerusalem, ifthey're still in business.  (Check with Ofer). 
Again:  we are speaking of SCAN to ASCII; Scan in graphics mode(ie, of handwritten transcripts) is of value only for archival,not editorial, purposes.
A Scanner, with best available OCR software, should be availableat secure central locations, presumably Yakar (Jerusalem) andKehilat Jacob (NYC).
 
HP Scanjet 3p, List Price $599, supplied with Calera WordScan OCRsoftware and HP AccuPage; Tel:  208-396-2551.

Canon IX-3010 List Price $569, supplied with Caere OmniScan OCRsoftware.	 Tel: 714--438-3000 

(Other scan softeare, Listed by vendor, no description:
Caere Omni-Page Win 2.0 ($100)
Caere Omni-Page Pro 5.0 Win ($500!) )

The review characterizes the HP Scanner as much easier to use, andthe Canon as providing more user control.  However, the reviewdoes not focus (!) on relative capability for scanning poorquality typed copy; which is the primary need in inputting R.Shlomo transcripts.                               

3.  VOICE RECOGNITION

Terminology: 
   Voice Recognition/ Dictation System / Continuous speech /
We are not concerned with 'Voice Navigators' nor (as far as Iunderstand, though I'll stand corrected by any programmer) with'Development Platforms') .  Nor with 'discrete speech' which is atype of crude real-time-only dictation program.  We are concernedwith a 'speaker-dependent' program, and do not necessarily need a'speaker-independent' program
    
Object:  to input tapes from a single speaker to disc as ASCII.
Most R. Shlomo tapes are likely to be of poor quality, withconsiderable "white" tape noise.  Volume of speaker's voice variesconsiderably, in some cases below level of tape noise.  Vocabularyis specialized but finite; only a few contexts.  Training can onlybe from tapes.  If necessary, a standard vocabulary could probablybe pulled out of tapes, for phoneme training.

PC Mag divides Voice-Recogntion systems into Dictation Systems,Voice Navigators (a gimmick for folks who would rather shoutorders at their computer than type on it), and DevelopmentPlatforms (for programmers who wish to add voice-recogntionfeatures to existing programs.)  So we're concerned only withDictation Systems.
	PC Mag notes that Dictation systems usually require"dedicated coprocessor cards".
	The review presupposes Windows or OS/2.

	PC Mag distinguishes between dication systems that require"discrete speech" (suitable only for business executives who aretoo lazy to type) and "continuous speech" .
	Note that in Voice-Recogntion dictation, unlike OCR Scanningof transcripts, Hebrew (input transliterated) is not a problem
    
	PC Mag notes that in general speaker-independent programs"still lie in the future".  It also notes that "the market ischanging and maturing rapidly."

	Training is usually by phonemes, which requires the user todictate a string of phonetic syllables into the program.  I do notknow if training programs by word exist.  One would expect so.  Ifso, training should be practical with R. Shlomo tapes, since thisis a single-user with a highly specialized but limited vocabulary,                            
	In general the problem with R. Shlomo tapes (unlike PVKtapes) is that the speaker did not in general speak directly intothe microphone, there is usually considerable background noise,and recording quality ranges from adequate to poor; one must,further, anticipate problems of tape deterioriation from age, poorstorage conditions, and, like,  orphaned nth-generation tapes ofunaffluent ancestrage.  (Eg a cheapo fast-dubbed copy of a cheapocopy of a lost original).                        

	







