The result is expected if you are processing and saving to PDF Text Under Image.  In this mode, by your specification, the software rasterizes (creates image of) every page in order to produce the output where page image is visible and text is stored under it.  This is the reason for decrease in quality, because the image visible in the document will be more pixelated than digital text, which is hidden under the image.  Also, because in the result PDF now there is a newly create picture of every entire page + the OCR text result, whereas before it was only digital text, the file size can increase substantially.  It takes more storage to store newly created images per page in the result PDF.

I tested a digitally-generated PDF file containing 10 pages and some color graphics.  I think this testing scenario will reproduce this common situation well.  No compression or down-sampling has been specified in export settings for this test, which if used in PDF export settings can help decrease file size further.

• Original digital 10-page PDF
  Contains digital text and some digital color graphics.
  30.5 KB
• Processed PDF, Text Only
  Digital test is visible, along with some OCR mistakes.  Formatting around graphics has been altered slightly.
  15.4 KB
• Processed PDF, Text Under Image
  Rasterized pixelated picture of each page is visible.  Perfect preservation of original look and formatting.  Text is stored under page pictures for selection and searching.
  763 KB
• Processed PDF, Text Over Image
  Rasterized pixelated picture of each page is visible in some graphics, with good preservation of original formatting, but the text is sharp due to being placed on top of page picture.  OCR inaccuracies are also visible.
  67.8 KB

Notice that this test applies only to ‘digitally-created’ PDFs where text already exists in vector form.  As a result, when saving to Text Under Picture, a whole new picture layer is created, which increases the storage size.  If you were to process ‘image-based’ PDF such as a scan, it would have contained the image of the page before processing.  OCR would add a text layer only, which is small in size, and the size difference would not be noticeable.

Selecting one of three available PDF export types will affect the look and size of the output PDF.  PDF Text Under Image is the most commonly used format for archiving, indexing, and preservation, but in some cases it comes at a cost of size increase if original PDF did not contain an image layer.  Available quality and compression algorithms included in the software can help decrease the output size.

The only known method how to clear that list is through Registry.

Navigate to \HKEY_CURRENT_USER\Software\ABBYY\FlexiCapture\9.0\Shell\MRU_List and remove unwanted items from the “List” node.

NOTE: Registry editing is dangerous if performed incorrectly.  Please do it at your own risk.  If unsure, copying original entries into Notepad so they could be restored if something happens during testing.

This issue is due to a potential bug in an installer, and it is easy to resolve. To resolve the issue, it is nessesary to create string EnableMSILog and set its value to “false” in HKEY_CURRENT_USER\Software\ABBYY\Debug branch of the Registry.  After the modificaiotn, the Registry screen should look like this:

Restarting the installation process should complete smoothly after this modification.

• to save a backup copy of the project in its current state for backup purposes
• to make a copy of the project before applying new changes, where a copy can be used as a recovery if needed
• to send or transfer project settings to a different installation , such as between development and production

The snapshot created includes global project settings, all templates and document definitions (and naturally their content), and optionally batches with documents.  If you intend to include batches with documents in your exported copy, select those batches when performing the export.  If you do not wan to have any batches included, make sure to not have any batches selected.

Open your Administration Station (in Standalone Edition) or Project Setup Station (in Distributed Edition) and do the following:

1. Go to Project menu, and select Export Project.
2. Browse to a desired destination, give it a meaningful name (we recommend to include date and time) and click Create.
3. A copy of the project will be saved in your destination.  It looks like a folder with sub-folders, but the entire folder is the entire project, so it should remain intact when moving or sending it.

Here are the two different behaviors of the .Items depending on the placement of the script rule:

1. The Script rule is stored in the Table or Column object (to which most developers are used to).  In this case the Column object has NO .Items collection.  Columns can be accessed individually by name as a regular field, but script acts on all cells at once.  The advantage of this script is simplicity since row iteration becomes unnecessary.  The disadvantage is that the Cells count is unknown, and we cannot work with individual cells.
2. The Script rule is stored elsewhere, but NOT in the Table of Column object.  In this case the Column object has .Items or .Items.Item(i) which returns a collection of Cells.  We can iterate through each cell or access individual cells.

The above is targeting working with Columns and collections of Cells within columns.  So in order to process Rows, we need to access an appropriate .Items.Item(i) of each Column where (i) represented the actual Row number.  Also, we must use approach # 2, which is to store our Script rule outside of Table or Column elements.  The most logical place to store such scripts is right in the Document Section page element (called “Page” in your Definition).

For example, let’s use some imaginary Table with multiple rows and columns and generate some rules.

a)      If Charges cell is populated and any other cells are empty in a row, then require Verification on those other fields

b)      If Charges cell is empty, then skip Verification on other fields.  (I even suggest to improve this logic by maybe clearing other fields from misc. data if Charges cell is empty)

Here is the VBScript code:

'======================================
' This Rule checks certain conditions in each row of GridLines
'======================================
dim i

'check that columns exist
'large all-inclusive IF-THEN with AND operator
if me.Field("FromDOS").IsMatched and _
    me.Field("ToTOS").IsMatched and _
    me.Field("POS").IsMatched and _
    me.Field("Units").IsMatched and _
    me.Field("CPT").IsMatched and _
    me.Field("DC").IsMatched and _
    me.Field("Charges").IsMatched Then

    'iterate through all rows, the below action within For loop happens per row
    For i = 0 to me.Field("Charges").Items.Count-1
        'check if CHARGES cell has at least one character
        'and other cellsFROM, TO, POS, CPT, DC, UNITS are empty
        If Len(me.Field("Charges").Items.Item(i).Text) > 0 and _
        Len(me.Field("FromDOS").Items.Item(i).Text) = 0 and _
        Len(me.Field("ToTOS").Items.Item(i).Text) = 0 and _
        Len(me.Field("POS").Items.Item(i).Text) = 0 and _
        Len(me.Field("CPT").Items.Item(i).Text) = 0 and _
        Len(me.Field("DC").Items.Item(i).Text) = 0 and _
        Len(me.Field("Units").Items.Item(i).Text) = 0 Then
            'make all fields as NeedVerification
            me.Field("FromDOS").Items.Item(i).NeedVerification = true
            me.Field("ToTOS").Items.Item(i).NeedVerification = true
            me.Field("POS").Items.Item(i).NeedVerification = true
            me.Field("CPT").Items.Item(i).NeedVerification = true
            me.Field("DC").Items.Item(i).NeedVerification = true
            me.Field("Units").Items.Item(i).NeedVerification = true
        End If

        'check if CHARGES has no characters
        'and irrelevant if other cells FROM, TO, POS, CPT, DC, UNITS contain data
        If Len(me.Field("Charges").Items.Item(i).Text) = 0 Then
            'make all fields as NOT NeedVerification
            '(Perhaps a better option here is to clear out all other fields)
            me.Field("FromDOS").Items.Item(i).NeedVerification = false
            me.Field("ToTOS").Items.Item(i).NeedVerification = false
            me.Field("POS").Items.Item(i).NeedVerification = false
            me.Field("CPT").Items.Item(i).NeedVerification = false
            me.Field("DC").Items.Item(i).NeedVerification = false
            me.Field("Units").Items.Item(i).NeedVerification = false
        End If        

    Next

End If


This code can be modified to perform any other logic while iterating through any number of Rows and calling on specific Columns by name to perform some actions.

For all of you out there with theoretical and practical experience, what is the best hardware and server types for large-volume OCR conversion.  What type of processors and other resources have you found most effective and why?  Please give a short description of your OCR work and loads to illustrate your feedback.  Thanks for your feedback.

Todd Gruff, B & G Manufacturing Inc.’s lead for FlexiCapture ingratiation, said WiseTrend solution and services were top notch.  “We were very impressed with ABBYY FlexiCapture’s ability to capture the data we required.  The software is easy to use and, with the terrific support from WiseTrend, the software’s ability to be adapted to different variations of Purchase Orders is limitless”.  B & G Manufacturing Inc. is a world-renowned manufacturer and supplier of machined parts.  The company receives high volumes of Purchase Orders that require accurate and efficient processing and data gets imported into SAP.

IBT, Inc. is a well-known wholesale industrial supplier interested to optimize and streamline Invoice processing and storage.  “After starting our data capture project with the industry’s standard approach, we switched to a more predictable approach using WiseTrend methodology.  After numerous testing this approach has proven to be the best way.  The single complex template would not produce reliable results on different variations.  There was too much time required filling in missed fields and even double checking all seemingly successful fields (aka false positives).  Using this new methodology we can control and rely on the data capture result.  Even when vendors are consistently changing their invoice format, which would cause serious complications and a major need for professional services in the past, using WiseTrend’s method it is quick and efficient to make the change.”  Kevin Thompson and Randy Bledsoe are accountants at IBT, Inc., who now run and maintain their own full data capture system without any IT involvement.  Captured data and images get exported to DocuWare.  The project was led by Toshiba Business Solutions, a major copier integrator and ABBYY VAR.

We are proud to announce our new method of servicing invoice processing and AP data capture automation needs.  WiseTrend WiseINVOICE implementation combines best modern technologies, proven methodology, and user-friendly training focusing on user self-sufficiency.  This solution is part of a growing trend to eliminate runaway costs for professional services while obtaining a reliable AP automation system through a truly customizable fully functional solution with immediate and accurately calculable ROI.

Up until now, meticulous programming or applying the “one-size-fits-all” approach for Invoice processing were only solutions in the industry.  This often misleading approach led customers to projects outside of budgets, skyrocketing professional service cost, long integration period, frustration, and inability to see clear return on the technology investment.  As the quantity of document variations increases linearly, the complexity of the system and associated setup increases exponentially.

With over 10 years of data capture experience, we developed a solution that works to optimize the technology and minimize production labor.  Our programmers previously spent weeks un-tangling projects that were based on the current industry lure—a one-size-fits-all solution for invoice processing.  Software manufacturers promise a single solution as a magic pill for all template variations.  Contrary to the promise, we spent many professional service days running regression testing on complex implementations for pre-setup AP projects.  Over many projects, we streamlined the Invoice and Purchase Order project setup to become a consistent and repeatable process.

Our approach provides tangibles that are surprisingly uncommon in the industry, including predictable timeframe for achieving customer’s return on investment as well as removing the often unpredictable run-away professional service costs and deadline extensions.  Following the proven formula, customers can anticipate the accurate calculation of time and effort it would require to build an effective data extraction system for a particular quantity of variations.

QUESTION
=================

I am extracting texts from OCRed Tiff files by using a library and dumping it in database. The text I am extracting are actually FORMS having fields like NAME, DOB, COUNTRY etc.   Since OCR does not know the difference between actual value and the label, it’s just dumping all text. Now I have text in DB in following format:

Name: MyName Address: My Address

Now the next step is to extract values lile MyName and MyAddrss from the DB. The document types may vary hence a generic parser might not work.

What would you suggest to do in this situation? Should I write different parsers? I am working on .NET

ANSWER
=================