Magnum One

home *** CD-ROM | disk | FTP | other *** search

/ Magnum One / Magnum One (Mid-American Digital) (Disc Manufacturing).iso / d11 / ocrshare.arc / OCRSHARE.DOC < prev next >

Wrap

Text File | 1991-02-24 | 129.3 KB | 2,930 lines

.pl OCRSHR22.ZIP OCRSHARE Release 2.2 Shareware User's Manual Solution Technology Inc. 1101 S. Rogers Circle, Bldg. 14 Boca Raton, Florida 33487 Tel. (407) 241-3210 Fax. (407) 241-3251 Copr (c) 1984-1990 Solution Technology, Inc. page 1 OCRSHR22.ZIP Table of Contents ================== Front Matter Copyright Acknowledgements....................... 1 License.......................................... 1 Bulletin Boards.................................. 2 Public Domain and Shareware Libraries............ 3 Association of Shareware Professionals (ASP)..... 4 Registration..................................... 4 Introduction......................................... 5 Why OCR Software?................................. 5 Features.......................................... 6 OCRSHARE...................................... 6 ATXSHARE...................................... 6 Advantex OCR.................................. 6 Ver 2.2 Enhancements.............................. 7 How to Use This Manual........................... 7 Tutorial...................................... 7 Trouble Shooting.............................. 7 Installation......................................... 8 Computer Requirements............................. 8 Running from a Hard Disk.......................... 8 Notation.......................................... 8 Hard Disk Installation............................ 9 Modifying CONFIG.SYS.............................. 9 Files & Buffers................................... 9 Starting OCRSHARE................................. 10 Forcing a Display Adaptor......................... 10 Screen Selections................................. 11 Mice.............................................. 11 Networks.......................................... 12 EMS Memory........................................ 12 Tutorial............................................. 15 Starting OCRSHARE................................. 15 Moving Around..................................... 15 The Six Basic Keyboard Keys................... 15 Short Cut Keys................................ 16 Exiting OCRSHARE.................................. 16 Perform OCR on an Image File...................... 17 Load a scanned image file..................... 17 Exploring the loaded image.................... 17 Load an OCR Recognition Font.................. 17 Setup the Text Output File.................... 17 OCR the full page............................. 17 Selecting areas to OCR or Cutout.............. 18 Canceling a selected area..................... 18 OCR of selected area.......................... 18 Unloading an OCR Font......................... 18 Saving a selected area........................ 18 Training a New OCR Font....................... 19 Saving the New OCR Font....................... 19 Copr (c) 1984-1990 Solution Technology, Inc. page 2 OCRSHR22.ZIP More about OCRSHARE's File Finders................ 19 Editing the File Name......................... 20 Over typing the File Name..................... 21 Incrementing Numbered File Names.............. 21 Graphics Editing.................................. 21 Invert Page................................... 22 Flip Page Vertical............................ 22 Erase Inside.................................. 22 Erase Outside................................. 22 OCR Notes............................................ 24 Looking at the output text file................... 24 Additional Information on Training................ 24 Naming the ink blots during training.......... 24 Proper training and symbol diversity.......... 25 Making an Alphabet Page....................... 25 Point sizes and Scanning DPI.................. 25 About Autoskip and Manual Training Modes...... 26 More on Autoskip Training..................... 26 About the Interactive Training screen......... 26 Training Your Own Fonts for Recognition....... 27 Using Your New OCR Font....................... 27 Multi font Capability......................... 27 Handling Special OCR Problems..................... 30 Broken Letters................................ 30 Broken Dot Matrix............................. 30 Run together Letters.......................... 30 Big Dirt Spots................................ 30 Underscores................................... 31 The % symbol.................................. 31 Italics....................................... 31 Advanced Training Methods......................... 31 Training Foreign Characters................... 31 Training from FRENCH.ATX...................... 32 Training Foreign Language Letters............. 32 Font Editing for correct OCR output........... 32 Using your new French OCR font................ 33 Training Special Symbols.......................... 33 Ocr Tuneup Fonts.................................. 33 Creating a tuneup font........................ 34 Derivative Fonts.................................. 34 Testing an Existing Font...................... 34 Editing the font.............................. 35 Retraining your font.......................... 35 Trouble Shooting..................................... 36 OCRSHARE V2.2 Registration Form...................... 44 Note To Retail Dealers............................... 45 Note To Shareware Dealers............................ 45 Copr (c) 1984-1990 Solution Technology, Inc. page 3 OCRSHR22.ZIP Copyright Acknowledgements ========================== MS-DOS is a registered trademark of Microsoft Corporation. IBM, PC-XT and PC-AT are registered trademarks of International Busi ness Machine Corporation. PC Paintbrush, PC Paintbrush +, and Publisher's Paintbrush are registered trademarks of ZSoft Corpo ration. PageMaker is a registered trademark of Aldus Corpora tion. Ventura Publisher is a registered trademark of Xerox Corporation. H.P. is a registered trademark of Hewlett-Packard Company. License ======= OCRSHARE (c) 1986 ,1987, 1988, 1989, 1990 is copyrighted software program product of Solution Technology, Inc.(STI), Boca Raton, Florida. OCRSHARE is being distributed under the shareware dis tribution process and is intended for the personal use and enjoy ment of the recipient. OCRSHARE is copyrighted and has been released for distribution as SHAREWARE. Please note that a great deal of effort and time has invested in the development of this program. You are granted a license to try OCRSHARE for a reasonable trial period without risk. 1. No organization, individual or other entity may reproduce, print, duplicate, copy, or distributed the program, manual or any ancillary file herein for any commercial purpose or personal gain or commercial gain whatsoever without the expressed written permission of Solution Technology, Inc. 2. STI hereby grants the user a single user license for the users private use and enjoyment, the programs, documents and Aancillary files herein. 2. The user is additionally granted the right to and encouraged to: a. print a hard copy version of this manual for his own use. b. give, copy, upload and otherwise distribute WITHOUT CHARGE the complete and full and true copy of the OCRSHRxx.ZIP file. 3. STi makes no warranty, either expressed or implied, with respect to the shareware software described herein, its quality, performance, marketability, or fitness for any particular purpose. Copr (c) 1984-1990 Solution Technology, Inc. page 1 OCRSHR22.ZIP 4. The user acknowledges that STi is not liable for any dam ages, either consequential or direct, arising out of the use of this Software. The user assumes complete responsibility for any decisions made or actions taken based on information obtained or distributed through use of this software product and any accompanying instructional or reference materials provided by Solution Technology, Inc. Bulletin Board License ---------------------- 1. Operators of electronic bulletin boards (Sysops) are hereby licensed and encouraged to post OCRSHARE for downloading by their users subject to the following provisions. In addition sysops are encouraged to keep the uploaded .ZIP file name OCRSHRxx.ZIP where xx represents this version of OCRSHARE (eg OCRSHR22.ZIP) so your users will know if they have the latest ver sion. 2. OCRSHARE may be uploaded to and downloaded from commer cial systems such as CompuServe, the Source, and BIX, so long as the only charge paid by the subscriber is for on-line time and there is no charge for the pro gram. Those copying, sharing, and/or electronically transmitting the program are required not to delete or modify the copyright notice and restrictive notices from the program or documentation; anyone doing so will be treated as a contributory copyright violator. 3. If you, as a BBS sysop or user, are passing this pro gram on others, uploading it to a bulletin board sys tem, or including it in a users group library, do not separate the files contained in the distribution ar chive - pass the entire archive on to the intended party. This ensures that those who receive the program will have all the correct configuration utilities and documentation necessary to get OCRSHARE up and running quickly. A listing of what files you should have and the purpose of each is listed later in this manual. 4. The OCRSHARE documentation may not be modified by users. The program may not be separated from the docu mentation when distributed. Printed or Photocopies ("Xeroxed") copies of the OCRSHARE documentation (i.e., this manual) may not be distributed or sold without the written permission of STI. 5. This license to use OCRSHARE does NOT include any right to distribute or sell OCRSHARE for commercial purposes, gain, compensation or profit. No entity or person other than Solution Technology, Inc. may accept payment or royalties for this program without an expressed written agreement with STI. Distribution terms are given below. Copr (c) 1984-1990 Solution Technology, Inc. page 2 OCRSHR22.ZIP Public Domain and Shareware Libraries ------------------------------------- 1. Distributors of "public domain" or user-supported software libraries must obtain written permission to distribute copies of OCRSHARE. No one may use OCRSHARE as a promotion for any commercial venture or as an enticement for the user to pay for any program, product, or service unless they have received the express written permission of Solution Technology, Inc. 2. You must obtain written permission from Solution Tech nology to distribute OCRSHARE. Please use the vendor application supplied near the end of this user manual. If you do not receive a reply, write again: our silence does NOT constitute permission, and you may not dis tribute, "pending" receipt of permission. 3. A maximum disk fee as set by Solution Technology in the above vendor contract must not be exceeded. OCRSHARE may not be included on any disk sold for more than this maximum. Major CD-ROM or optical disk libraries are exempt from this restriction, provided that they have STI's permission to distribute OCRSHARE. 4. Vendors may not modify or delete ANY files on the disk. Vendors may add a "GO" program, and/or a reasonable number of small text files designed to assist or pro vide a service to the user, but these added files must be easily identifiable and end-users must be allowed to delete the added files. 5. Vendors must make a reasonable effort to distribute only the most recent versions of OCRSHARE. All vendors who have requested and received written permission to distribute OCRSHARE will receive new MAJOR releases as they are issued. 6. All disk vendors must comply with any and all vendor guidelines vendor requirements set forth by the Associ ation of Shareware Professionals (ASP); for more infor mation about ASP, contact its chairman, Jim Button, at Buttonware in Seattle. Violation of any ASP guideline or requirement automatically revokes permission to distribute OCRSHARE. Until formal requirements are adopted by the ASP, you must comply with the following guidelines: Vendors must make an attempt to educate users on the nature of Shareware. Catalogs, advertisements, order forms, and all disks sold should contain ASP-approved or recommended wording describing the nature of share ware, and should explicitly state that no part of disk Copr (c) 1984-1990 Solution Technology, Inc. page 3 OCRSHR22.ZIP sale revenues are paid to the programs' authors. When vendor catalogs or advertisements carry both Shareware and PD programs, the Shareware programs must be differ entiated from the public domain programs in some way (in the description, with an asterisk, by listing the registration fee, etc.). Association of Shareware Professionals (ASP) -------------------------------------------- OCRSHARE is a Shareware program conforming to standards as established by the Association of Shareware Professionals (ASP) located at 325 118th Ave. S.E., Suite 200, Belleview, WA 98005. Registration ------------ If you find OCRSHARE useful you are encouraged to register you copy. The base registration fee is $45. With registra tion you will recieve in return the latest version of OCR SHARE called ATXSHARE which does not have the shareware nags and allows graphics translations between .PCX, IMG and .TIF image files. Additionally registration allows you to obtain a full set of pretrained Advantex OCR fonts for only $25. You will find a registration form at the end of this docu ment which will aid you in registering your copy of OCR SHARE. Please keep in mind that we must have a registration form on file for you before we can offer product support. Copr (c) 1984-1990 Solution Technology, Inc. page 4 OCRSHR22.ZIP Introduction ============ OCRSHARE is a complete shareware version of STI's Advantex OCR (Optical Character Recognition) package which is de signed to convert scanned image files into text and picture files for use by other programs in your computer. After using OCRSHARE to input a document, you can manipulate the extracted data using your favorite word processor, desk top publishing, or graphics programs. OCRSHARE is smart, fast and flexible but still easy to use. We're sure you will find OCRSHARE an indispensable tool to reduce those endless hours at the keyboard attempting to computerize volumes of printed material. Why OCR Software? ----------------- When it comes to inputting existing documents and hard copy data into your computer, you basically have only two options: either enter the data "manually" by retyping it into your system, or have an Optical Character Recognition (OCR) system do it for you. The page image is first scanned into a bitmapped image file using the utilities provided with your scanner. Each image file represents a single page and contains many rows of black and white dots (called pixels) which come from detect ing the light reflected from the characters, lines, logos, dirt and other ink patterns on the page. While these graphic images look like characters on your graphics display, they are still only collections of dots. OCRSHARE's major purpose is to learn to identify these ink patterns and output its ASCII equivalent into a text file. For example, a graphic "A" becomes an ASCII "A", and so on until each character has been translated or converted to text. OCRSHARE can read using its dictionarys of existing symbols (stored in .FTF files) or easily learn to recognize new symbols, company logos or characters in other accented alphabets such as German, Greek, Spanish, French and Rus sian. Thus graphic symbols and characters are translated into their letter equivalents. Note that this process is letter and word recognition, not language translation. OCRSHARE can even be taught to read hand printed letters if the penmanship is neat and consistent. Copr (c) 1984-1990 Solution Technology, Inc. page 5 OCRSHR22.ZIP Features -------- Features vary from product to product as follows OCRSHARE Features ----------------- - Reads scanned images TIF, PCX and IMG files - Output ASCII, Wordstar or Word Perfect files - OCR Features - Trainable OCR Fonts - Handles Skewed Text - Reads Monospaced, Proportional and Typeset Text - Comes with Helvetica, Times Roman & Courier OCR Fonts - Adjustable Automatic Dirt/Spotting Filter - Adjustable Automatic Graphics/Line/Box Filter - Image Cleanup features - Erase Inside - Erase Outside - Rotate Page - No Direct Scanner Support ATXSHARE Features ($45 Registeration Required) ------------------------------------------ - All of the OCRSHARE Features above - Latest Version - No Shareware Nags - Save/Cutout TIF, PCX, IMG Files (eg file conversion) _ Optional OCR font library ($25) AdvanTex OCR ($395) ------------------- - Direct Scanner Support for Cannon, Microtek, HP Scanjet and ScanJet Plus, Panasonic RS505,506, Chinon 2000 and 3000, Ricoh, Mitsubishi full page Handscanner, and others. - Batch Mode Support - Multipage Support - Save/Cutout TIF, PCX, IMG Files (eg file conversion) - Ocr Font Library included - Other extensions. Copr (c) 1984-1990 Solution Technology, Inc. page 6 OCRSHR22.ZIP Ver 2.2 Enhancements -------------------- The following improvemets, enhancements, features and capa bilities were added to the AdvanTex Release 2.2 base program in upgrade to earlier versions. - Added a line separation preprocessor(Edit Menu) for text which has lines joined vertically by a few touch ing letters. - Added a ligatured (joined character) error detector which properly marks untrained ligatures in the output text (with XX). - Added proper " processing - Added word break test processor(Edit Menu) - Improved context rules for capitalization and numbers - Improved recognition at 200 dpi - Improved recognition rules for all recognition - Improved and expanded tolerance for line skew - Improved word space processing. - Upgraded all Scanner Drivers - Upgraded all OCR Fonts How to Use This Manual ====================== Most PC users, novice to expert, generally hate to read manuals. Right?! That's why we've made OCRSHARE as intuitive and easy to use as possible. We have included easy-to-understand lessons in our tutorial section. These lessons provide you with clear, step-by-step instructions on using all the simple, yet powerful features of OCRSHARE. Even if you are an expert computer user, you should at least glance through the lessons to get acquainted with OCRSHARE. Once you understand the basics, you can then use the manual's index to quickly find out how to make OCRSHARE do what you want it to do. Lessons ------- You learn even more about OCRSHARE by visiting the Tutorial section and going through the step-by-step examples in the Lessons. Trouble Shooting ---------------- Help is always available while running OCRSHARE. Simply press "Alt" and "H" together and a Help menu will appear on your screen. Copr (c) 1984-1990 Solution Technology, Inc. page 7 OCRSHR22.ZIP Installation ============ Computer Requirements --------------------- Your computer should have at least the following capabili ties. - An 8088, 8086, 80286 or 80386 based IBM/PC or compatible computer. The faster the CPU chip the better. An 8088/4.77 mhz will work, but will be unbearably slow. - DOS 3.0 or higher operating system. - A full 640 K of main memory. - A hard disk drive with at least 2-3 megabytes of free space. - Display card/monitor with graphics capability (Hercules monochrome, CGA, EGA, VGA, etc.) Running from Hard Disk or Ram Disk ================================== OCRSHARE can be installed in virtually any directory on any hard disk drive of your system, even copied and run from a RAM disk (most effective when running the disk EMS simula tor). Notation -------- - When you see [sp] the space indicated is mandatory so press the space bar on the keyboard. - When you see [enter] press the key marked ENTER on your keyboard. - When you see a letter preceded by the word CTRL (as in [CTRL-C], hold down the key marked "CTRL" or "Control" on your keyboard and simultaneously press the letter key indicated (in this case 'C'. Copr (c) 1984-1990 Solution Technology, Inc. page 8 OCRSHR22.ZIP Hard Disk Installation ---------------------- A summary of the installation steps are listed below. Installing OCRSHARE is quite simple and mostly involves copying files. 1. Create a subdirectory called OCRSHARE for OCRSHARE on your hard disk. You may use another directory name, although the examples assume OCRSHARE is used. 2. Use PKUNZIP to expand the files of OCRSHRxx.ZIP into the directory. 3. Modify CONFIG.SYS if necessary. 4. Log into the OCRSHARE directory and run the program by typing OCRSHARE[enter] Modifying CONFIG.SYS ==================== You should have a CONFIG.SYS file in your root directory. CONFIG.SYS is a file which is read by the computer during the power up boot and CTRL-ALT-DEL three-key boot sequence. CONFIG.SYS, which must be on the boot disk, contains configuration information for DOS as well as specifications for the loadable device drivers. While your CONFIG.SYS may contain more statements, the statements used by OCRSHARE are: FILES = <number of simultaneous files permitted> BUFFERS = <number of disk buffers available> DEVICE = <EMS memory card driver filename> FILES & BUFFERS --------------- We recommend the following minimum values for the FILES and BUFFERS statements in CONFIG.SYS. More that the minimum is perfectly OK. FILES = 20 BUFFERS = 20 Copr (c) 1984-1990 Solution Technology, Inc. page 9 OCRSHR22.ZIP Note: If there are multiple FILES and or BUFFERS lines in your CONFIG.SYS (usually put there by different application installation programs) keep only the largest and delete all other lines. Starting OCRSHARE ----------------- To start OCRSHARE type OCRSHARE[enter] OCRSHARE will normally automatically select the proper internal graphics display driver and display mode for your graphics card. Note: Color monitors are supported in black and white (2 color) modes only since color modes are too memory intensive for large images. Forcing a Display Adaptor ------------------------- If you have dual screens, an unknown brand display card, or desire a different display mode, you may have to override OCRSHARE's auto selection. To do this, simply add the appro priate switch parameter after OCRSHARE on the command line when you start OCRSHARE. For Example, The following forces OCRSHARE to use the 2 color 640X350 EGA driver. OCRSHARE /E:6 Upon entering this command the following message will be displayed: IBM Enhanced Graphics Adapter, 640 x 350 2-color to quit -press CTRL-C to continue -press any other key If you signal continue, OCRSHARE will set up for the card you specified specified. The hardware configuration is saved in the file OCRSHARE.CFG so that the next time you start OCRSHARE your display adaptor is automatically select ed. If OCRSHARE seems to hang on start up, you may still have an incompatible display card. Press CTRL-X to get back to the prompt or reboot and append the correct option switch for your specific display adaptor. Copr (c) 1984-1990 Solution Technology, Inc. page 10 OCRSHR22.ZIP OCRSHARE /E:6 [enter] if you have an EGA compatible card OCRSHARE /C:1 [enter] if you have a CGA compatible card OCRSHARE /E:7 [enter] if you have a VGA compatible card or OCRSHARE /xxx [enter] (where xxx is one of the settings below) Screen Selections ----------------- The screen selections supported at the time of the printing of this manual are listed below. Option Graphics Card Screen Size /A AT&T Graphics Adapter 640x400 /A:1 AT&T Graphics Adapter 640x400 /C:1 IBM Color Graphics Adapter (CGA) 640x200 /E:1 IBM Enhanced Graphics Adapter (EGA) 640x350 /E:6 IBM Enhanced Graphics Adapter (EGA) 640x350 /E:7 IBM PS/2 MultiColor Graphics Array(MCGA) 640x480 /G MDS Genius Display Adapter 736x1008 /H Hercules/AST Monochrome 720x348 /O Toshiba 3100 640x400 /T Tecmar Graphics Adapter 720x352 /T:1 Tecmar Graphics Adapter 720x352 /T:2 Tecmar Graphics-Adapter 720x704 /S STB GraphicsPlus-II Adapter 640x352 /S:1 STB GraphicsPlus-II Adapter 640x352 /S:5 STB GraphicsPlus-III Adapter 640x400 /W Wyse WY-700 640x400 /W:1 Wyse WY-700 640x400 /W:2 Wyse WY-700 800x400 /W:3 Wyse WY-700 1280x800 /X IBM 3270 PC 720x350 /X:1 IBM 3270 PC 720x350 For Compaq Monochrome Graphics Monitors - Use the following settings /A:1 Portable II and III plasma 640x400 /C Portable or CGA compatible 640x200 /E:1 EGA with monochrome screen 640x350 Note: VGA will always default to MCGA (/E:7) Mice ---- Although OCRSHARE does not at this time support a mouse directly, it will work if you have a programmable mouse hooked up to your computer. Such a mouse can emulate the directional keys of the keyboard (left, right arrow, etc.). Copr (c) 1984-1990 Solution Technology, Inc. page 11 OCRSHR22.ZIP Networks -------- OCRSHARE is not generally network aware but can be used from a server. Secondly if your network drivers occupy too much room in the first 640K OCRSHARE will not have suffient room to run. Also be aware of OCRSHARE's use of EMS memory with particular attention to server disk caching. To install simply copy all of the programs and files to a server direc tory. EMS Memory ---------- EMS memory provides up to 8 megabytes of additional memory to any IBM/PC compatible computer. This memory is not directly addressable by computer; it must be paged in and out to a 64 Kilobyte buffer by software calls to an EMM device driver. This driver, installed in CONFIG.SYS, must conform to the Lotus, Intel, Microsoft (LIM) specification 3.3 or 4.0. OCRSHARE and EMS memory ----------------------- OCRSHARE uses the EMM driver to access the EMS memory. OCRSHARE stores the full sized display image and the zoomed out image in the EMS memory. While you do not need external EMS memory for occasional work, you will find the perform ance advantages of a real EMS card significant if you are trying to process large numbers of pages. EMS Memory Requirements ----------------------- OCRSHARE images require memory real or simulated EMS memory. The chart below gives you some idea of how much memory you will need for different sized images. Image Resolution 8.5x11 in 200 Dpi about 500K 300 Dpi about 1.1 Meg 400 Dpi about 1.8 Meg The formula for calculating the number of bytes required is: (Resolution x Resolution x length x width) / 8 (bits per byte). Note: If you have insufficient real EMS memory to store the size of the image you are scanning, you will get an EMS allocation error. You can still process the image by setting OCRSHARE to simulate the EMS environment on disk, exiting and restarting OCRSHARE. Copr (c) 1984-1990 Solution Technology, Inc. page 12 OCRSHR22.ZIP Types of EMS Memory ------------------- Depending on whether you are using a real EMS card, an extended memory EMS simulator, or a disk EMS simulator, the memory used by the EMS system will, respectively, be paged in and out to a real EMS card, or to extended memory, or to your hard disk. OCRSHARE's internal EMS simulator is a highly optimized form of a disk EMS simulator. OCRSHARE's EMS simulator ---------------------------- If you do not have real EMS memory active, OCRSHARE will use the free space on your hard disk as temporary virtual memory and simulate the EMS function internally. If, for example, you are running OCRSHARE from drive C: which has, let's say, 400K of hard disk space remaining (use CHKDSK to determine), then there will not be enough space to run OCRSHARE. If you have another hard drive (e.g., D:) with more space in it, you can log into that drive (D:), and run OCRSHARE as long as OCRSHARE.EXE is in the path. Although OCRSHARE is installed on C:, because it is called from another drive (D:), the remaining space on that drive (D:) will be used as the simulator. Note: Make sure your hard disk has 1-2 megabytes of free space on it. You can use DOS's CHKDSK utility to see how much disk space is available. The OCRSHARE simulator is in effect while OCRSHARE is running. When OCRSHARE exits, all of OCRSHARE's simulator files are deleted. Installing a disk cache in EMS Memory ------------------------------------- SOME DELAYED WRITE DISK CACHE PROGRAMS, WHEN INSTALLED TO USE EMS MEMORY, DO NOT SHARE THE EMS RESOURCE PROPERLY WITH OTHER PROGRAMS AND CAN CRASH YOUR HARD DISK. If you try to run ANY EMS application with an "ill- behaved" delayed write cache program installed in EMS, it will likely result in a serious disk crash with corrupted FAT tables. "Well-behaved" EMS disk caching programs will properly save and restore EMS information as it operates. While cache programs usually operate properly when set to WRITE THROUGH mode (where all writes are done immediately), the problem occurs when the cache program is set to do DELAYED WRITES to the disk. This requires that the cache use a clock tick to interrupt the application program to do the disk writes. If the cache program DOES NOT properly save and restore EMS page information during this interrupt, it WILL cause a disk crash. Copr (c) 1984-1990 Solution Technology, Inc. page 13 OCRSHR22.ZIP This warning does not apply to WRITE THROUGH cache programs or VDISK programs which are generally well behaved. If you want to use DELAYED WRITES with your disk cache program, install the cache memory buffer Extended memory. Besides using the unused 384K that is typically there, this leaves the EMS Expanded Memory free for OCRSHARE to use. Scanner Installation -------------------- OCRSHARE and ATXSHARE are shareware versions of Solution Technology's Advantex OCR product which directly supports most desktop scanners. Due to the fact that the of scanners is often complicated and (and costly) in terms of telephone support these shareware products are not distributed with scanner drivers or scanning capability. To get direct scanner support you must purchase regular AdvanTex. (See the file README.1ST). OCR on prescanned Images ------------------------ You must scan images using your scanners built in utility and save them in a Tiff, PCX or IMG image file format. OCRSHARE can then load these images and perform OCR or basic image manipulation. Copr (c) 1984-1990 Solution Technology, Inc. page 14 OCRSHR22.ZIP Tutorial ======== With OCRSHARE, you are only a few keystrokes from obtaining usable text output in the form of either an ASCII text file from pages scanned and saved in PCX, TIF or IMG image file formats. This Tutorial is designed to get you up and running with OCRSHARE in the shortest possible time. If you feel that you need more detailed instructions, or you are going to use OCRSHARE almost immediately to do production work, simply begin with the regular Tutorial (Tutorial) Section. Starting OCRSHARE ================= 1. Start OCRSHARE by typing OCRSHARE at the DOS prompt, then press [enter]. Note: As long as OCRSHARE.EXE is in the PATH command, OCRSHARE can be called from any drive or sub-directory. Please refer to your DOS manual for more information about the use of PATH. Moving Around ============= To move the selector bar, use the up, down, left, and right arrow keys on your numeric/arrow key pad. The [PgUp] and [Home] key will move the cursor to the top of a menu, and the [PgDn] and [End] keys will move the cursor to the bottom of a menu. The Six Basic Keyboard Keys --------------------------- You normally only need to use the following six keys to execute virtually any function in OCRSHARE. Arrow Keys - The up and down arrow keys move the menu cursor bar up and down any menu. The left and right arrow keys move the cursor bar from menu to menu. Escape Key - The Esc (escape) key will acts as both your dismiss menu (or cancel) key, and as your attention key. By pressing the Esc key, OCRSHARE will stop what it is doing and return to the previous menu screen. [enter] Key - The [enter] key is used to select, execute or activate the action highlighted by the cursor. Once you have placed the cursor on the action of choice, press [enter] to activate that action. The [enter] key may be identified as RETURN or as an arrow on your keyboard. Copr (c) 1984-1990 Solution Technology, Inc. page 15 OCRSHR22.ZIP Short Cut Keys -------------- The following keys allow you to bypass moving around the menus. Home Key - This key moves you to the top of any menu or list. Home Key - This key moves you to the bottom of any menu or list. PgUp/PgDn - These keys move you up or down one list page on a scrollable list. If the list is not scrollable these keys act the same as the Home/End keys. CTRL Keys - A control key shortcut is indicated on the menus by a ^ symbol followed by a letter. To execute the function associated with a control key, hold down the CTRL key and press the letter indicated on the menu. Function Keys - A function key shortcut is indicated on the menus by a F and followed by a number. To execute the function associated with an F key, simply press the indicated function key on your keyboard. CTRL PgDn - When drawing a capture box with Select Area F9, this combination will move the cursor box down the page in much larger steps than the down arrow will. CTRL PgUp When drawing a capture box with Select Area F9, this combination will move the cursor box down the page in much larger steps than the up arrow will. CTRL -> When drawing a capture box with Select Area F9, this combination will move the cursor box to the right in much larger steps than just the right arrow will. CTRL <- When drawing a capture box with Select Area F9, this combination will move the cursor box to the left in much larger steps than just the left arrow will. Exiting OCRSHARE ================ When it is time to exit OCRSHARE, use the arrow keys to position the cursor over the Exit to DOS command under the INFO main menu. Press [enter]. You may also exit OCRSHARE by pressing the CTRL-X control key shortcut. Copr (c) 1984-1990 Solution Technology, Inc. page 16 OCRSHR22.ZIP Perform OCR on an Image File ============================ Load a scanned image file ------------------------- 1. Press F5 to select the image file loader 2. Press [enter] to select the OCRSHARE file format. 3. Press the Down Arrow Key to move the selector bar over the PAGE1.ATX file name 4. Press [enter] Once to select that file to load. 5. Press [enter] Again to load the selected file. Exploring the loaded image -------------------------- 1. Load an image file as described above 2. Select "Select Area" or press F9. A cross hair cursor will appear. 3. Use the Arrow Keys to move the cross hair cursor to interesting parts of the scanned image. 4. Press F10 to "Zoom In" on selected area 5. Press F10 again to "Zoom Out" to the full page view. 6. Press [esc] key to cancel the select area. Load an OCR Recognition Font ---------------------------- 1. Select "Font Settings..." on the OCR menu, or press CTRL-F. 2. Select "Load Existing Font". 3. Move the cursor bar so that it is over TMSROMAN.FTF in fonts list. 4. Press [enter] once to select the font. 5. Press [enter] a second time to load the font. 6. Press Esc to get back to the OCR main menu. Setup the Text Output file -------------------------- 1. Select "Text Settings..." on the OCR menu, or press CTRL-T. 2. Select "Set Output Text File". 3. Type DEMO at the File: prompt (the .TXT extension will be appended for you); press [enter]. 4. Press [Esc] to get back to the OCR main menu. OCR the full page ----------------- 1. Load one or more OCR Fonts. 2. Load a scanned page image file (PAGE1.ATX for tutorial). 3. Setup text output file name. 4. Select "Convert to Text" or press F4 to perform OCR on the full page and put the results into a newly created file called DEMO.TXT. This output text file can be sent to the printer, viewed from DOS, or imported into your favorite word processor. Copr (c) 1984-1990 Solution Technology, Inc. page 17 OCRSHR22.ZIP Selecting areas to OCR or Cutout ------------------------------------------ 1. Press F9 (Select area) 2. Move the cross hairs with the arrow keys. 3. Anchor the first corner by pressing the [enter] key. 4. Pull the tiny icon box (dragon) using the arrow keys until it surrounds the area desired. 5. Use the "+" and "-" keys to move the dragon to different corners of the larger capture box. 6. When you are satisfied with the positions of all corners, lock the capture box in place by pressing the [enter] key. Canceling a selected area ------------------------- 1. Press F9 (Select area) 2. Press [enter] twice. 3. Since the select area rectangle has no width or height it is now canceled and full page is implied. Ocr of selected area -------------------- 1. Mark an area of interest as described above. 2. Press F4 to perform OCR on the selected area and either append to or overwrite the existing file called DEMO.TXT. This output text file can be sent to the printer, viewed from DOS, or imported into your favorite word processor. Unloading an OCR Font --------------------- 1. Select "Font Settings..." on the OCR menu or press the CTRL-F key. 2. Use Down Arrow to move the selector to TMSROMAN.FTF. 3. Press [enter] to pop up the font management menu. 4. Press [enter] to remove TMSROMAN.FTF Saving a selected area ---------------------- 1. Mark an area of interest as described above. 2. Select "Save Area" or press F8. 3. Select a graphics format to save to. (OCRSHARE is OCRSHARE's fast load internal graphics file format. You must register to obtain the other formats listed). 4. Type a file name, say CUTOUT, at the File: prompt; the appropriate extension will be appended for you. 5. Press the [enter] key to use this file name. 6. The selected area of the image is now saved into the file CUTOUT.ATX, in the graphics file format you se lected. Copr (c) 1984-1990 Solution Technology, Inc. page 18 OCRSHR22.ZIP Training a New OCR Font ----------------------- 1. If you have not loaded PAGE1.ATX do so now as described above. 2. Use Select Area (F9) to draw a box around the letters. Make sure that you have some upper case or tall letters on each text line in the selected area. 3. Select "Font Settings..." or press CTRL-F. 4. If TMSROMAN.FTF is still loaded, unload it as described above. 5. Select "Make a New Font" and type in OCRFONT (or any other name you want). 6. Press the [enter] key to register that font for training. 7. Press [Esc] to get back to the OCR main menu. 8. Select "Train Font" or press F7 to begin training characters from the selected area into the newly created OCRFONT.FTF. a) As each letter is displayed and boxed, type the corresponding letter on the keyboard followed by [enter] key. b) If multiple letters are boxed, you can type them all in sequence, and press the [enter] key. c) If the symbol boxed has no equivalent keyboard letter, you should make up an unambiguous two or three letter sequence to name the letter, and press the [enter] key. Note that the next time that symbol appears for training you MUST use the SAME name you entered earlier. d) To skip over dirt, spots or badly broken letters, press [enter] without typing the name of a symbol. Only named symbols are added to the OCR font data base. e) Press the [Esc] key any time to stop training and get back to the OCR menu. Saving the New OCR Font ----------------------- 5. When you are finished with training of the new font Select "Font Settings..." or press CTRL-F from the OCR menu. a) Select OCRFONT.FTF (eg the ocr font you have just trained). b) Move the selector bar to "Save Font" and press [enter]. Except for the letters OCRSHARE hasn't seen yet, you have just created a partially- trained font. More about OCRSHARE's File Finders ================================== OCRSHARE uses a menu window called a File Finder to select for loading and name files for saving. A File Finder dis plays a list of the available files of a given type in a the Copr (c) 1984-1990 Solution Technology, Inc. page 19 OCRSHR22.ZIP currently specified directory. Explore this by the following exercise: 1. The cursor bar is initially located on the File: prompt bar. Press the up arrow key until the cursor bar is on the PATH: *.ATX prompt line. 2. Press the down arrow key until the cursor bar is in the box listing the OCRSHARE image file names. 3. Use the down arrow key again to move the cursor bar cursor bar down until it covers the file named PAGE1.ATX. 4. Press the [enter] key. Note that the file name and the cursor bar pop up to the File: prompt line. Press the [enter] key again to load the PAGE1.ATX file into display memory. While loading any file, OCRSHARE first displays a load progress window, then a Computing Full Page View progress window as it constructs the full page view of the image. When loading has finished, OCRSHARE will display the full page view on your monitor. 5. Press the ESC key to hide the OCR drop down menu as illustrated above. This lets you view the page without anything in the way. 6. Press the ESC key again and the OCR main menu will reappear. Editing the File Name --------------------- All OCRSHARE text fields have an easy to use, built in, line text editor. Notice the small vertical bar to the right of PAGE1.ATX. This means that the field editor is insert mode. 1. Press the [Ins] key several times to toggle between insert and overstrike mode. In insert mode the cursor indicates the point between letters where the next character will go. In overstrike mode the block cursor covers the character which will be overwritten. It indicates that you can edit the characters of the file name in this field. 2. Press the Home key to move to the beginning of the text field. 3. Press the End key to move to the end of the text field. 4. Press Left arrow key to move the text cursor (vertical bar) to the point between the 1 and the period. This is the position where we are going to begin editing. 5. Press the Backspace key to delete the 1. (You could have also moved to the point between the E and the 1 and pressed the Del key). 6. Type the number 1 to reinsert the number 1. Copr (c) 1984-1990 Solution Technology, Inc. page 20 OCRSHR22.ZIP 7. Type the letter a to change PAGE1.ATX to PAGE1a.ATX 8. Press the Backspace key to delete the a. Over typing the File Name ------------------------- You can also type over the existing file name. 1. On any finder menu, move the selector bar to the File: field. 2. Press the Home key. The text cursor will move to the far left side of the file name. 3. Press [Ins] (insert) key until the text cursor changes to the type-over fode. OCRSHARE is ready to type over the existing charac ter or the entire file name. 4. Type in the filenames letters and press [del] key to delete any trailing letters. 5. Press [enter] to use this filename and the path speci fied by the Path: field. Incrementing Numbered File Names --------------------------------- OCRSHARE is designed to work easily with filenames that end in numbers. Since PAGE1.ATX ends with the number 1 to make it a numbered file. Note that the file extension (in this case .ATX) indicates the type of file information rather than just serving as the end of the filename. 1. Press the F2 function key on your key board. The file name will change to PAGE2.ATX. 2. Press the F10 function key and the File name changes back to PAGE1.ATX. You may increment or decrement the file name in this manner for all numbers between 0 (zero) and 9999. Notes: 1. Each type of finder has a separate data storage areas for their current information. This means that you can set the path for the OCRSHARE finder to one directory and the path of the PCX finder to another directory and OCRSHARE will keep them straight. Graphics Editing ================ The following graphics menu performs editing functions that modify pixels, areas or the entire page in the display memory. This enables you to modify scanned images to improve the quality of graphics and OCR projects. OCRSHARE can load and save images to or from PCX, TIFF, or (our own) OCRSHARE file formats thus allowing you to convert image files for different purposes. Copr (c) 1984-1990 Solution Technology, Inc. page 21 OCRSHR22.ZIP Invert Page ----------- Invert Page which makes negatives of positive (black on white) images. 1. If you don't have the image PAGE1.ATX .ATX still loaded, load it again. 2. Select Invert Page by moving the cursor bar over Invert Page and pressing [enter]. 3. The black and white image reverses itself. 4. Press [enter] again and the image reverses, bringing it back to its original state. It is that easy. Flip Page Vertical ------------------ Flip page vertical rights pages which were accidentally scanned upside down. 1. Select Flip Page Vertical and the image will turn upside down so that the top is now the bottom. 2. Press [enter] again and the image returns to its original upright position. Erase Inside ------------ Erase Inside erases the area INSIDE a "Select Area" rectangle and is good for selectively eliminating unwanted parts of the scanned page. The area selected can be as small as one bit or as large as the entire page. 1. Use Select Area to enclose an area inside the monitor screen in the image as shown above. 2. Select Erase Inside. OCRSHARE will erase everything within the box. The monitor in the image now appears clear with a black border around it as shown above. Erase Outside ------------- Erase Outside erases the area OUTSIDE a "Select Area" rectangle and is perfect for eliminating edge trash. The area selected can be as small as one bit or as large as the entire page. 1. Use Select Area to enclose an area inside the monitor screen in the image as shown above. 2. Select Erase Outside. OCRSHARE will erase everything BUT what is inside the select area box. The monitor in the image now appears clear with a black border around it as shown above. Notes: Copr (c) 1984-1990 Solution Technology, Inc. page 22 OCRSHR22.ZIP 1. Your original image, PAGE1.ATX, has not been modified, and will not be until you elect to save over it. Copr (c) 1984-1990 Solution Technology, Inc. page 23 OCRSHR22.ZIP OCR Notes ========= 1. Due to memory constraints, we provide direct, internal, support for only a few word processor formats at this time. If you are going into a different word processor, you should select the Wordstar document format, and the import utility provided with your word processor to convert the ASCII text file into your word processor's internal text format. 2. OCRSHARE automatically avoids small spots of dirt as well as pictures in PAGE1.ATX. 3. You may occasionally see extra spaces or symbols seemingly run together in the Progress Monitor window. In addition you may even see two or three lines as OCRSHARE wraps the text output on the display. Don't fret, the monitor window is only an early preview of how conversion is doing. The actual page formatting is done later, after OCRSHARE can analyze the relationship of ALL of the symbols from the page. Therefore, the actual text (.TXT) file will show the most accurate translation. Looking at the output text file. ------------------------------- You must leave OCRSHARE to examine your output text file PAGE1.TXT. 1. Press CTRL-X or select Exit to DOS on the INFO main menu. 2. At the DOS prompt type in the following command: TYPE PAGE1.TXT [enter] Additional Information on Training ================================== Naming the ink blots during training ------------------------------------ Font training is the process by which the user teaches OCRSHARE to recognize new characters. In its simplest form, training is a two-step process. 1. OCRSHARE automatically locates and draws a box around an unknown ink pattern. 2. You type in one or more ASCII character(s) to name the ink pattern. OCRSHARE will remember your choice, and during OCR translation it will output these ASCII character when it detects another ink pattern closely approximating the one trained. Copr (c) 1984-1990 Solution Technology, Inc. page 24 OCRSHR22.ZIP Proper training and symbol diversity ------------------------------------ One of the major components of properly training a new OCR Font is to select an area of symbols from a scanned page of text that contains the font desired and a good mixture of upper case letters, lower case letters, numbers and punctua tion. Making an Alphabet Page ----------------------- We have found that the following letter sequence, taken from a standard keyboard does quite a nice job of training up a character set. Not perfect, mind you, but easy to implement and use. Note that there are extra spaces between each symbol. Note also that the punctuation MUST NOT be put on separate lines otherwise its size and relative position relative to the rest of the letters in the font will be lost. ` 1 2 3 4 5 6 7 8 9 0 - = \ ~ ! @ # $ % ^ & * ( ) + A a B b C c D d E e F f G g H h I i J j K k L l M m N n O o P p Q q R r S s T t U u V v W w X x Y y Z z , . < > / ; ' : " [ ] { } Since we have found that it is usually sufficient to train each symbol 3 to 8 times, you should, for convenience, put a number of sets of the above sequence on the same page. In addition, you should generate a test paragraph (Quick brown fox paragraph) which uses all of the letters and punctuation in normal words and sentences. Point sizes and Scanning DPI ---------------------------- We suggest using 14 point type when printing an Alphabet Page using a publishing system and doing your scanning at the highest available resolution to get the best possible typeface rendering. Since OCRSHARE has OMNISIZE we are training on a typeface, not a point size. Note: You need to press [character] [enter] to tell OCR SHARE's memory. If [enter] alone is pressed, no training for the current symbol is accomplished. Note: Should you inadvertently strike the wrong key(s), you may untrain the mistrained symbol. See Untrain Symbol described later for details. Copr (c) 1984-1990 Solution Technology, Inc. page 25 OCRSHR22.ZIP About Autoskip and Manual Training Modes ---------------------------------------- As shipped OCRSHARE comes up in Manual Training mode. This means that it stops on every symbol encountered. If you switch to AutoSkip training mode OCRSHARE will, when at least the minimum number of symbols have been trained, stop only on symbols which do not meet minimum acceptable criteria for matching symbols currently in the OCR font data base being trained. More on Autoskip Training ------------------------- Autoskip means that OCRSHARE will automatically skip over any character during training that meets the following default criteria: The menu fields within the parentheses below are found in the Convert Settings Control Panel menu which can be accessed by pressing the [CTRL-C] key. - The training count for a character is at least 2 (per Min Training Count) - The confidence level is between 0-40 (per Confidence Threshold) - The confidence level of the second choice is at least 15 points more than the first choice (per Similarity Threshold) Autoskip will allow you to train a new font quickly and accurately by ignoring known characters and concentrating on unknown or unclear characters. About the Interactive Training screen. -------------------------------------- Understanding this information is very important to properly train a font. The categories are as follows: Train Count: Indicates the number of times OCRSHARE has been trained on the particular character it is display ing in its Closest Choice List. You rarely need to train OCRSHARE on any single character more than 3 times. Confidence: OCRSHARE is numerically telling you how confident it is that the Closest Choice List characters match the character it has drawn a box around at the top of the screen. A good match falls in the range between 0-40. A number higher than 100 indicates that OCRSHARE has no confidence it recognizes the boxed character at the top of the screen properly. Copr (c) 1984-1990 Solution Technology, Inc. page 26 OCRSHR22.ZIP In Autoskip training mode, as the training progresses, OCRSHARE will start to see the characters for a second time, and the confidence rises (numbers start to drop dramatically) as the correct closest choices are displayed. Font Name: Indicates what font in memory OCRSHARE has chosen its closest choice list characters from. This indicator is only important when more than one font has been loaded in memory. Closest Choice List: Indicates the best matches OCRSHARE has for the boxed character at the top of the screen. The Closest Choice characters always appear in descend ing order with the best choice at the top of the list, the fourth best at the bottom of the list. Trained Alphabet: The list of characters in this box indicate to you which characters have been trained at least one time, and what characters OCRSHARE has yet to see. Every time you enter a new character into OCR SHARE's memory, it will darken that character on the Trained Alphabet list. Characters OCRSHARE has yet to see will remain halftone. An effectively trained font will have at least all upper case letters (A- Z), lower case letters (a-z), and punctuation marks darkened. Translation Progress Window: This window will give you an indication as to where the translation is in relation ship to the scanned page. The translation Progress Window is NOT to scale. Do not be concerned if the spacing of characters and words is not exact, or if extra carriage returns appear on the screen. Training Your Own Fonts for Recognition --------------------------------------- Although we have provided 3 pre-trained fonts, you may find that you get best results when you train your own pages for recognition. This is because there are so many fonts in existence and therefore we cannot train them all. Using Your New OCR Font ----------------------- You can use NEWFONT alone or in conjunction with other existing fonts simply by loading into memory using the "Font Settings..." function menu. Multi font Capability -------------------- It is important to note that OCRSHARE can utilize more than one font at a time. Multiple font capability will enable you to properly translate a very broad range of printed material containing mixed fonts on the same page. Copr (c) 1984-1990 Solution Technology, Inc. page 27 OCRSHR22.ZIP Note: Very small type or low DPI scanning resolution may require you to train special fonts as the height of the characters may approach the lower mathematical limits of OCRSHARE's OMNISIZE capabilities. About the Convert Settings Control Panel ---------------------------------------- The options found on the Conversion Control Panel allow you to modify the operation of OCRSHARE's OCR engine and identi cally affect both training a font and translating the char acter images to text. Normally you will not have to change any of these settings. Zone - This parameter determines which area of the image will be trained, or converted to text. "Whole page" will cause the training or conversion to start at the left topmost character on the page and will proceed to the right bottom character. "Select area" will cause training to begin with the top left most character inside a box drawn by Select Area and will proceed to the right bottom character in the capture box. "Page mask" behaves in the same way as Select Area, except that it uses a Select Area which remains fixed, page after page. This is very useful when converting text from the same zone on many pages. Training Mode - Allows you select one of two modes for a font, Manual Training and Autoskip training. Autoskip will be the mode that is used most often once you gain familiarity with OCRSHARE. "Autoskip" Autoskip training will occur at an accelerated pace, because as OCRSHARE becomes more confident of the symbols you are teaching it will stop to train only characters in which it has lower confidence. "Train OCR" Manual training stops on every symbol, regardless of the confidence level. This allows for training of characters up to 128 times each (optimal training is 3-8 times per character). View Translation - Translation Monitor ON allows you to view the conversion to text. Translation Monitor off causes a faster translation to occur since the text does not have to be displayed. NOTE: The actual text file is created at the very end of the viewing stage, when the message Formatting Output Page is displayed. Copr (c) 1984-1990 Solution Technology, Inc. page 28 OCRSHR22.ZIP Set Max Symbol Size - Allows you to determine the largest character or graphic that OCRSHARE will recognize during training and translation. When this function is selected, the dimensions of the capture box (drawn with Select Area) will automatically be placed in Edit Max Symbol Size setting . Any graphic larger than this box will be ignored. This is extremely handy for processing pages containing both text and graphics, since the graphics will be ignored during OCR. Set Page Mask Bounds - This will determine the area of the page mask, or the area that will be recognized during training and translation. When this function is se lected, the dimensions of the capture box (drawn with Select Area) will automatically be placed in Edit Page Mask Bounds. Once the Training mode is set to Page Mask, OCRSHARE will deal only with images inside this area. Furthermore, only this area will be active from page to page while Page mask is selected. Set Spot Filter Size - Determines the size of the graphic below which is ignored during training and translation. For example, any graphic less than 15 pixels in size will be ignored. The shape of the graphic does not effect the filter in any way. Edit Max Symbol Size - This feature allows you to edit, the symbol size by entering the number of pixels you wish the length and width of the symbols to be. However, this function will rarely be used since the dimensions shown here are usually set by Set Max Symbol Size. Edit Page Mask Bounds - This feature allows you to edit the page mask by entering the number of pixels for each edge of the capture area. However, this function will rarely be used since the dimensions shown here are usually set by Set Page Mask Bounds. Min Training Count - This number, between 2 and 9 inclusive, determines the minimum number of times each character will be trained before OCRSHARE will consider skipping over it during training. Max Training Count - This number, between 2 and 9 inclusive, determines the maximum number of times any character is trained when Training Mode is set to autoskip. Confidence Threshold - This number only has effect in the Manual Train, Autoskip mode and does not usually need to be changed. OCRSHARE's certainty in identifying characters depends on a number of parameters, including the quality of the scan, the consistency and quality of the type and how well the font is trained. Changing the Confidence Threshold number determines how sure we want OCRSHARE to be in identifying characters before auto- skipping to the next character. Setting the confidence Copr (c) 1984-1990 Solution Technology, Inc. page 29 OCRSHR22.ZIP threshold to 0, however, means that OCRSHARE will NEVER autoskip except for PERFECT matches. A Confidence Threshold of 40 is represents allowing about 15% error before a good match is failed during training. Similarity Threshold - This number only has effect in the Manual Train, Autoskip mode and does not usually need to be changed. Changing the Similarity Threshold deter mines how close the first two choices of symbols must be before OCRSHARE will stop and ask for confirmation. OCRSHARE wants to have confidence that it is correctly differentiating its choices before it skips to the next letter. Setting the Similarity Threshold to 0 effec tively turns the detector off. Handling Special OCR Problems ============================= Broken Letters -------------- You can lose proper symbol capture if letters are broken in such a way that a clear vertical white space is visible. This is usually caused by a light ribbon or an nth order photocopy. First, try rescanning using your utility soft ware with a darker contrast setting. Secondly, reprint the page with a new ribbon. Third, use manual training and skip over the broken letters. Broken Dot Matrix ----------------- See Broken Letters (above). Run together Letters -------------------- OCRSHARE can separate letters which overlap if there is at least a thin white space snaking down between them. OCR SHARE will, however, clump multiple symbols together into a single symbol if they physically touch or run togeth er in any manner no matter how thin the touch point. First, try scanning with a lighter contrast setting or a higher Dpi resolution. Secondly, reprint the script training page with a space between each letter. Third, use the Erase Inside function with a one pixel wide box to erase a thin line between the joined letters. Fourth, use manual training and either skip over the run together letters or train them as letter pairs. Big Dirt Spots --------------- If the page has dirt spots larger than the minimum spot size because these spots might be treated as a valid symbol. First, try cleaning the scanner glass. Secondly, inspect your printed training page and use white out to cover any spots you may find (these are usually flecks in the paper itself). Third, try scanning with a lighter contrast set Copr (c) 1984-1990 Solution Technology, Inc. page 30 OCRSHR22.ZIP ting to reduce sensitivity to spotting. Finally, use the Erase Inside and Erase Outside function to eliminate any problem areas. Underscores ----------- On some printers the underscore symbol is so far under the normal baseline that it often becomes separated into a separate line. If this is occurring, it may cause an occa sional spurious symbol to appear on a line by itself. The % Symbol ------------ In some fonts, OCRSHARE will (sometimes) separate the % symbol into a lower case o followed by a /o. If this is occurring, you may have to train using the Manual training modes. We suggest tagging the lower case o as %o and the /o as %/o then using the font editor, redefining %o as the ignore symbol (~~) and %/o as the % symbol. Italics ------- For reasons similar to the % symbol, in some italic fonts, OCRSHARE set the dots above i's and j's and the dots below ?'s and !'s. If this is occurring, you should train the page using a manual training mode. We suggest tagging and redefi nition in a manner similar to that used for the % symbol above. Advanced Training Methods ========================= While there are many printed symbols which do not appear on your keyboard, you will quite often find that you want to recognize them and translate them into something sensible in your output text file. Training Foreign Characters --------------------------- Foreign alphabets have accented symbols which are not found in the English alphabet and, as a consequence, do not occur on US keyboards. Handling foreign letters and special symbols requires two steps. - Train the foreign font database, coding the special characters as described. - Edit the font database, redefining all of the specially coded characters so that it will, when performing OCR, form the correct letters and output them to your output text file. Copr (c) 1984-1990 Solution Technology, Inc. page 31 OCRSHR22.ZIP Training from FRENCH.ATX. ------------------------ 1. Load the FRENCH.ATX image file. 2. Use Font Settings CTRL-F to define a new font called FRENCH.FTF. 3. Select Train Font F7, or press F7. 4. Train normal letters with single keystrokes. When you encounter a foreign language symbol, refer the to the next section to learn how to type in a two-character sequence for the symbol. Continue until you have completed training the page. Note that the double letter symbols are NOT displayed in the alphabet window. Training Foreign Language Letters --------------------------------- As you train, you will need to name the letters of the various foreign alphabets which may not appear on your keyboard. When this occurs, you will have two choices. The first is to input the proper ASCII extended character number by depressing the ALT key and typing the number on your number key pad. A chart with all of the appropriate ASCII extended character numbers has been supplied in the Trouble Shooting section of this manual. There is an alternative method. You can make up an equivalent training name for each of these unusual symbols. The following examples show some of the cases we have encountered and the practical coding rules we made up to train these foreign letters on an English keyboard. Font Editing for correct OCR output ----------------------------------- When you are finished training from FRENCH.ATX, you need to edit how the font outputs the foreign language symbols to the text file. Select Font Settings... CTRL-F from the OCR main menu. 1. Select the "Font Settings..." sub menu 2. Move the cursor bar down to FRENCH.FTF 3. Move the cursor bar down to Edit and press [enter]. 4. Pressing the [end] key. 5. Move the cursor bar to the first multi letter symbol you want to redefine (e') and select it by pressing [enter]. 6. Select Redefine from List. 7. Move the cursor down the list with the down arrow key until it rests on the forward accented e symbol. Press [enter] to select this e'. 8. Repeat steps 5,6, and 7 for each of our specially coded French letter. 9. Press [esc] when you are finished to get back to the "Font Settings..." menu 10. Save your newly modified font Copr (c) 1984-1990 Solution Technology, Inc. page 32 OCRSHR22.ZIP Using your new French OCR font ------------------------------ Provided you have captured all of the letters, numbers and punctuation of the typeface, you are now ready to use the OCR function (F4) to convert French documents printed in that typeface. Let's try it now on the same FRENCH.ATX that we have loaded. Set an output file using Text Settings... CTRL-T. Press F4 to activate the conversion. Training Special Symbols ======================== OCRSHARE's ability to recognize special symbols can be quite handy when you intelligently redefine them for your particular needs. In this section you will learn how we code and redefine a few special symbols in ways that generate useful output. This section is intended to be used as an example as, in reality, there are thousands of potential uses for this technique. Copyright symbol Code as (c) Bullets We suggest that you should code square bullets as "[-]" and round bullets as "(-)". For example, in Ventura Publisher text we use the font editor to redefine both of our training keystrokes "(-)" and "[-]" to the same string "@BULLET = ". The Hand Symbol. If you want to train OCRSHARE to recognize the Hand symbol we suggest that you code this as "!H!". For Ventura Publisher we redefine "!H!" to "@NOTE =" OCR Tuneup Fonts ================ Occasionally you will use an existing font which does not give you a good translation. Often the reason is a variation in the design of the printed typeface. For example, there are over 25 significant variations of the Courier typewriter font. Other frequent causes are letter distortions caused by poor quality printing or nth order photocopying. Another frequent cause is the accidental ligatures generated by either scanning at too low a resolution or by the original typesetter setting the letters so closely that they physically touch. In either case this causes a new, untrained, symbol to be defined. Copr (c) 1984-1990 Solution Technology, Inc. page 33 OCRSHR22.ZIP The easiest way to handle this is to train a tune-up OCR font which will augment your regular font. A tune-up font is loaded in addition to your base font to supply the additional trained symbols you need for this particular document. Some tune-up fonts you create will be temporary, meant only to aid the translation of a specific problem document. These can be discarded after you use them. Others tune-up fonts will be variations of a base OCR font. These you should keep to load again with your base font. Creating a tune up font. ------------------------ Note: You can press the ESC key at any time to back out of these menus without doing anything. 1. Select the Font Settings...CTRL-F menu. 2. Select Load Existing Font. Select and load your base font, e.g., TIMES.FTF. 3. Select Create New Font. [enter] TIMES-HP. This will create the tune up OCR font TIMES-HP.FTF. 4. Load or scan the page containing the font to be tuned. 5. Interactively train the whole page or a selected area of the page where the problem symbols are. (OCRSHARE will only stop on symbols which are problematical in BOTH the base and the tuneup OCR databases). Note: type in multi character strings when OCRSHARE stops on an run together characters. 6. When completed, save the tune-up font in the normal manner. 7. To use the tune up later, load the base font AND the tune-up font and proceed normally with your OCR trans lation. Derivative Fonts ================ Occasionally you will use an existing font which does not give you a good translation. When the primary reason is a variation in the design of the printed typeface, as an alternative to tuning up a font, you may want to either correct the existing font or create a derivative font. In either case the procedure is identical. Testing an Existing Font. ------------------------- 1. Set up the Convert Settings control panel for normal interactive training. 2. Load the OCR base font which seems to be giving prob lems. 3. Load or scan the sample page which is giving problems when using the specified base font. Copr (c) 1984-1990 Solution Technology, Inc. page 34 OCRSHR22.ZIP 4. Perform a test translation. Write down the symbols which are not being recognized properly. Editing the font. ---------------- 5. You now need to untrain each of the problem characters in the OCR font. Bring up the Font Editor by selecting Font Settings...CTRL-F, then highlighting the font and pressing [enter], and selecting Edit. 6. Use the down arrow key to move to each problem charac ter noted in the test translation. 7. Select it. 8. Select Untrain Symbol. The definition of that symbol will be deleted from that fonts database in memory. It is now as if the symbol was never trained in the first place. The font database on disk has not been modified because we haven't saved the one in memory yet. 9. Repeat steps 7,8 and 9 for each problem character. Retraining your font. -------------------- 10. Using the newly edited font AND the same image, activate Train Font F7. OCRSHARE will stop on each of the symbols you've untrained (and occasionally some of the other normal letters) so that you can retrain, or enhance the training of already-trained symbols. This should go very quickly. 11. Once you've finished training the font, you should save it. If you save the font under a different name, you have created a derivative font. If you save the font under its original name, you have corrected the exist ing font. Copr (c) 1984-1990 Solution Technology, Inc. page 35 OCRSHR22.ZIP Trouble Shooting ================ The following is a checklist of items to verify before running OCRSHARE. I installed OCRSHARE according to the instructions, but when I started it, I did not get the menus display. a) The first thing to try is delete the OCRSHARE.CFG file if it exists then restart OCRSHARE. If the graphics display adaptor shown when you restart is not the one you have installed then you will have to override OCR SHARE's selection. b) If the above fails, check to see if the monitor and display card you have is really a graphics display. If it is, make sure that you are selecting the correct override switch for the graphics display card you have installed. c) Delete the OCRSHARE.PRO setup file. The machine locks when I try to run OCRSHARE a) Be sure enough memory is available, at least 520K free. Use CHKDSK to be sure. Also, release all TSR's (termi nate and big resident programs) you may have loaded before running OCRSHARE, we need the space. b) Make sure that a proper screen driver is loaded (see Selecting a Screen Driver in Installation. The machine just jumps back to DOS when I run OCRSHARE. This may be because you are using the EMS simulator and have very little space left on your hard disk. Be sure to make 2- 4 megabytes available on your hard disk, or run from another hard disk or partition with more space on it. The menu display is up but the keyboard won't work. Press the NUM LOCK key; the NUM LOCK light on the keyboard should always be OFF. The Mouse Won't Work OCRSHARE does NOT support a mouse at this time unless your mouse is programmable in such a way that it will emulate keystrokes such as up arrow, down arrow, enter, Page Down, etc. Consult your mouse manual for specific instructions for setting up your mouse in this manner. Copr (c) 1984-1990 Solution Technology, Inc. page 36 OCRSHR22.ZIP When saving an image, the hard disk light "churns forever", and the on screen status bar barely moves. There is not enough hard disk space available to which to save the image. Press Esc, and either use the File Manager (INFO menu) to delete some files and make more room, or Exit to DOS CTRL-X to do the same. The scanned image is garbage, or no image shows at all. The most likely problem here is that you are using a scanner which has its own image buffer mapped into the address space above 640K and another device such as an graphics display card, an EMS card or a network card which conflicts with it. Verify that an EMS conflict is or is not occurring by selecting Load Page F5 and loading the EMS test page. If the cross hair pattern is not clean has the same garbage as the scanned image, there is an EMS conflict. Resolve the address conflict by changing the address of the scanner card or the address of the conflicting card. Make sure that you edit the CONFIG.SYS file to tell the drivers. Finally, make sure that the scanner is clean. You can easily check to see if the unit is clean by scanning a pure white page. If an image with many spots results, clean the glass. The Quick Start procedure shows nothing/garbage in the Translation Window. Your scanner is probably not working properly. Press Scan Page F3 to scan the WELCOME TO OCRSHARE page. An image of the page should show up on your monitor in black characters on white background. If not, walk through the Scanner Problems, and Scanning sections earlier in Trouble shooting. When training, nothing happens. If nothing happens, and you are "kicked" back to the main menus, make sure that the loaded image consists of dark characters on light background. If not you must invert the page using "Invert Page" off the EDIT menu. OCRSHARE is skipping characters that it has never seen before. This is probably because the untrained character being skipped is too much like another that has been trained. a) Press Esc to exit the training screen. Copr (c) 1984-1990 Solution Technology, Inc. page 37 OCRSHR22.ZIP b) Select Convert Settings... CTRL-C and set the Training Mode to Train OCR manual. c) Use Select Area F9 to capture the word(s) in which the character(s) being skipped are located. d) Select Train Font F7 and train the character(s) that were being skipped. e) When finished, you may reset the Training mode to Train OCR, autoskip. I keep seeing "Out of Memory at row XXX. Please set narrower window". If your page is very complex, you made need to select an area of text to train/convert on, rather than the whole page. More often than not, the cause is skewed, or sloping, lines of text. Use Straighten Page on the GRAPHICS menu to horizontally align the image. Also, perform a CHKDSK from DOS to verify that at least 520K is free before running OCRSHARE. Be sure to release any TSR's before running OCRSHARE. How do I know which font to load when converting to text? A font guide comes with the registered version of OCRSHARE. If you have this guide. Compare the typestyles on your page to those of the samples and load the one(s) that are the most similar. If you are not sure which font is most like yours, you can load up to five fonts that are most like the one on your page. There are too many/too few line feeds and or/spaces in the text file. a) View the text (.TXT) file from DOS using the TYPE command, or use your favorite word processor. The Translation Window in OCRSHARE is not to scale, meaning not all spacing is shown correctly. b) Remember to set Proportional line Spacing if the type on the page is proportionally spaced, and to set Fixed line spacing if the type is fixed-spaced or typewritten material. How do I skip a characters during training? Press [enter] only without naming the symbol. A symbol is trained only when character(s) are entered and return is pressed. Copr (c) 1984-1990 Solution Technology, Inc. page 38 OCRSHR22.ZIP What happens if I try to train a font when more than one font is loaded? OCRSHARE will show you the "best match" for the current character based on all the loaded fonts, even though only the font set to "Train" is being modified. Unless you are training an existing font as a derivative font (see Lesson 5), you should remove all fonts other than the one being trained for best results. Then simply reload all the desired fonts for conversion to text after training is finished. When F7 (Train Font is selected) nothing happens, and the program returns to the menus. a) Make sure that the image is dark characters on light background. If not, change the Image Background (Scanner settings... CTRL-S) to the opposite of what is currently displayed, and re-scan the page. b) Make sure that Whole page is the Conversion zone (in Convert Settings... CTRL-C) selected. Or, that the Select area or Page mask are of sufficient size to capture the area of the page you want to train on. The characters are badly broken or OCRSHARE is stopping on pieces of a character during training. 1. Try increasing the scanner contrast to a higher setting (toward Dark) so that the characters darken up and the broken pieces become continuous (see Lesson 1). 2. Use the following "Tag" coding of the character as to ignore or consolidate parts of the broken symbols. a. "Tag" the first piece of the character by entering a two- or three- character sequence, then press return. b. Tag the second piece of the character by entering the correct symbol. For example, if the character is supposed to be an "A", but appears in two pieces, identify the first piece as "AA", and the second as "A". c. When training is completed, prepare to Edit the font per the instructions under Editing the Font. d. Once the Font list window is displayed, Page Down to the end of the table and locate the two- and three- character symbols. Highlight the row with "AA" and press return. e. Select Redefine as Text and enter the ignore marker, i.e., two tilde marks (~~) . f. Repeat step 5 for each of the remaining multi- character tags. g. During conversion to text, the first piece will now be recognized, but nothing will be output; the second piece will be translated to the symbol assigned to it. Copr (c) 1984-1990 Solution Technology, Inc. page 39 OCRSHR22.ZIP More than one character is included in the training box on the screen. a) Try setting the scanner resolution to its highest DPI setting. This insures that the scanner has the best chance of seeing white space (if any) between adjacent characters. (see Lesson 2). b) If the scanner is already at its highest DPI setting and you still have excessive problems with run together letters, try decreasing the scanner contrast a little bit towards lighter. This insures that the scanner has the best chance of seeing white space (if any) between adjacent characters. c) Some characters may be actually run together. Train these groups of run together characters as a multi- letter (ligatured) symbols. You will usually find that the typesetter consistently ran certain letter pairs together. As long as the sequence is 40 characters or less, the corresponding sequence can be entered. For example, if the sequence "iff" should appear in the training window instead of "i", "f", and "f", enter "iff" from the keyboard, then press return. The dots on the i's and j's are not included in the box during training. Try setting the Spot Filter (in Convert Settings... CTRL-C) from 15 to 8 dots. OCRSHARE is not finding the dot since its size is smaller than the one specified in the Spot Filter setting. More than one character, vertically, or more than one full line, are being displayed during training. a) Occasionally, more than one line of text will be so closely spaced that descending characters (e.g. "y") will hang into the next line of text. In these instances, OCRSHARE cannot determine where one character ends and the other begins, so it will capture both as one character. Do not train on any such symbol, just press return only. b) Another possibility is that two lines have some dirt, malformed or particularly large characters invading the interline space which causes a lack of line separation. The trick here is to provide OCRSHARE with just enough space between the lines of text so that it can distinguish among them. A very simple method for doing this is to draw a box (using Select Area, F9) between the two lines whose dimensions are the length of the line horizontally, and about 1/32 to 1/16 inch in height (a very short rectangular box is needed). Once the box is drawn, select Erase Inside (GRAPHICS menu) and all marks within this box causing poor line separation will be removed. Copr (c) 1984-1990 Solution Technology, Inc. page 40 OCRSHR22.ZIP Some lines of text are accurately translated, yet others are completely mis-translated; or, characters from two consecutive lines are appearing in the training window. a) This usually happens when OCRSHARE is having trouble with line separation. That is, OCRSHARE cannot distinguish the bottom of one line from the top of the next. Check the physical page to make sure there are no stray marks between the lines that are displayed simultaneously. Also check for any underlined words. Try erasing or masking these marks and re-scan the page. b) Make sure the image is straight. If it is not, re-scan the page, or Straighten the image (see Lesson 1). The larger characters, and sometimes normal-sized characters, are being ignored during training and/or translation. Remember that any symbol larger than the Max Symbol Size setting (Convert Settings...) will be ignored. This will include any large single characters and any multi-character symbols (e.g. ARL) whose combined size is larger than the specified symbol size. Even though the characters A,R,L are normal-sized, their being so close makes them appear as one character to OCRSHARE. Can OCRSHARE handle columnar material? Yes. It is best to convert columns to text one at a time by using Select area and drawing a box around each column one at a time. What is the best way to OCR a complex page, i.e. one with both pictures and text on it? There are actually three ways a complex page can be processed: a) Perhaps the best way is to do nothing, as OCRSHARE will automatically ignore graphics (pictures) as it performs OCR on the page b) Select Convert to Layers (GRAPHICS menu). This feature will separate the image into three files. The first, LARGE.RAS will contain all images (usually pictures) greater than the Max Symbol Size (Convert Settings...). MEDIUM.RAS will contain all the images (usually text images) greater than the spot filter setting and less than or equal to the Max Symbol Size. SMALL.RAS will contain most of the stray marks and spots on the page. Converting to layers is a convenient way of automatically separating the page into its various components. Copr (c) 1984-1990 Solution Technology, Inc. page 41 OCRSHR22.ZIP c) Use Erase Inside on picture portions of the page. If you do not wish to keep the pictures, you may decide to remove them entirely form the image so that you can see what you are doing with the text more readily. Can OCRSHARE handle landscape-oriented material At this time, OCRSHARE does not have a 180-degree rotate function, but will be in the future. Can OCRSHARE run under Windows? Yes. Conditionally. OCRSHARE is NOT a Microsoft Windows compatible program. Either run OCRSHARE without a .PIF file and use all of Windows defaults or set up a .PIF file which gives OCRSHARE the MAXIMUM amount of memory and indicate that it directly modifies the screen and keyboard. Keep in mind that OCRSHARE, like Windows, is a large program and thus will probably run best as a stand alone process. Can fonts trained at one resolution be used to translate text scanned at another? Yes, but the accuracy can be lower as compared to the accuracy when using a font trained at the same resolution as the scanned page. Which TIFF format does OCRSHARE read and write to? Our TIFF format is essentially the same as the ones generat ed and read by Hewlett-Packard Scan Gallery software and have been tested against a number of other software pack ages. How do you train if more than one font is present on a page? It is best to train all the characters of one font and save it to a file, remove this font, and then to train all the characters of another font and save this training in another file. Try not to mix characters of different fonts during training of any given font. Remember that Select Area (F9) can be used to capture various words and lines of text to augment the training process should characters of one font be intermixed with characters of another font. Can OCRSHARE be trained to read quotation marks? No. Each piece of the mark must be assigned a ' since OCRSHARE captures only one ink pattern at a time. Copr (c) 1984-1990 Solution Technology, Inc. page 42 OCRSHR22.ZIP Can more than one font be trained at the same time? No. If more than one font is loaded (Font Settings), then only one o these can be set to Train mode at any given time; the rest are set to Use. How many bytes of (EMS) memory do scanned images require? It depends on the resolution, but the number can be deter mined by simple calculation: At a resolution of 200 Dpi, there are 200 bits by 200 bits, or 40,000 bits of informa tion per square inch). Therefore, the formula for determin ing the number of bytes required is (Resolution x Resolution x length x width)/8 bits per byte. On a 8 1/2 x 11 page, then, there are (200 x 200 x 8 1/2 x 11) / 8 or 467,500 bytes of memory required to store this image. Copr (c) 1984-1990 Solution Technology, Inc. page 43 OCRSHR22.ZIP OCRSHARE V2.2 Registration Form =============================== Name: ------------------------------------------------------ Address: ___________________________________________________ ___________________________________________________ City:_______________________________________________________ State(or Provence+Country):_________________________________ Zip/Postal Code:_____________ Phone: _______________________ **** YOU MUST BE REGISTERED TO QUALIFY FOR UPGRADES **** Qty ____ Registration(s) of OCRSHARE $45.00 each $__________ ____ OCR Font Disks $25.00 each $__________ ____ Advantex OCR $395.00 each $__________ ATXSHARE/Advantex Media Selection [ ] 5.25 DSDD (640K) [ ] 5.25 DSHD (1.2Meg) [ ] 3.5 DSDD (720K) TOTAL $__________ Make Check or money order should be made payable to, UPS COD availible on request: Solution Technology, Inc PO Box 273372 Boca Raton, Florida 33487 Copr (c) 1984-1990 Solution Technology, Inc. page 44 OCRSHR22.ZIP NOTE TO RETAIL DEALERS: ======================= For more product information on OCRSHARE, ATXSHARE or Advan text Retail Dealers should contact Solution Technology, Inc. directly at the address and phone number given on the front cover of this manual. NOTE TO SHAREWARE DEALERS: ========================== Shareware Dealers must contact STI directly to apply for distribution. Any request for shareware catalog inclusion requires that you send a current copy of your product cata log with your application, along with a blank diskette and mailer. Once you have been approved, you will receive written permission from Solution Technology to include OCRSHARE in your library for distribution. Additionally, you will receive your diskette back containing the latest OCRSHARE release. You must, in addition send each issue of your catalog as they are published to update our mailing list of active vendors. STI's continued receipt of your catalog will both enable us to verify that you have the most recent OCRSHARE release (within reason), and that your business is still active. We will send authorized shareware dealers each MAJOR revi sion of OCRSHARE as it is released so that your library is kept up to date. Copr (c) 1984-1990 Solution Technology, Inc. page 45