Separation and Separation Review

From Grooper Wiki
Jump to navigation Jump to search

Separation uses various techniques to group pages into classified documents.

This is an example of the Separation Review Attended Client interface.

About Separation and Separation Review

Grooper uses various approaches and algorithms to determine the classification of a page or folder. The various settings on a Content Model and Document Type really add to the complexities for separating pages into documents. ESP Auto Separation removes but does not eliminate a lot of the manual work to separate and classify documents. Separation Review is a new review module designed to make the manual work quick and easy.

Use Cases

Separation

Training Scope

Adjusting the Training Scope provides benefits to the accuracy and performance of ESP Auto Separation by focusing what is important when it comes time to separate and classify Unstructured paginated documents. For example, the Normal mode will create a single Form Type and divide trained examples into "First", "Middle" and "Last" pages. From individual document to individual document, often the most meaningful features composing them are found on the first and last pages, and there can be more variance on the pages in between. This is different from the previous approach, which created individual Form Types for each trained example, each with their own "Page X of X" Page Type objects. This unifies all trained examples into a single Form Type, making the training and classification of these documents ultimately simpler and more efficient. The FirstLast mode assumes meaningful features for classification are only found on the first and last pages, with the middle pages containing no information needed to make a separation or classification decision. With this mode enabled, only trained examples of the first and last page and their associated features will be saved. This can improve processing time by removing all the features in the middle pages for consideration. The FirstOnly mode narrows this scope even further by only storing features from the first page of trained documents.

Writer's Note: While these properties are available on Structured, Fixed, and Extended Pagination types, these Training Scopes are almost entirely pertinent to the Unstructured Pagination type. If you use these Training Scopes on anything besides an Unstructured Document Type, you may see they always are trained according to the "Normal" scope.

Normal - This is the classic version of capturing training features for a document type.

  • This mode differs slightly for Unstructured Pagination. Instead of trained middle pages (ie Page 2 of 4 and Page 3 of 4) created as individual PageType objects of their multipage Form Type parents, they are created as "Middle" Page Type objects of a single Form Type. Furthermore last pages (ie 4 of 4 or 7 of 7) are combined into a single "Last" Page Type object.
  • This cleans up the Node Tree by creating a single unstructured Form Type for documents of various page lengths, instead of many. This will speed up runtime classification and can potentially yield more accurate results.

FirstLast - This is handy for training only the first and last page of a document type. It lowers the features training requirements, improves speed, and allows middle pages between the first and last page to be combined with the first and last.

  • Note: This approach assumes the middle pages will not be classified as confident first pages of a Document Type. For ESP Auto Separation. When a page is classified as the first page of a document, it a high priority indicator as the start of a new document.

FirstOnly - If someone has used "Document Titles" extraction with a Positive Extractor for separation in the past, consider this property an upgrade to that approach.

  • A common approach to document classification is to write a positive rule extractor locating the document's title on the first page of the document. However, this approach can break down when an image is poor quality, causing the Document Type's title extractor to miss the title.
  • The FirstOnly mode can allow the continued capture of features in titles combined with other trained features on the page. That way, if the title is missed, the separation engine can fall back on other features from trained first pages of the Document Type.
  • This approach is highly similar to the Extended Pagination in concept, where the first page is used as the separation point and all pages after are included in the document folder until another confident first page is found. This approach differs in that only the first page is trained, eliminating the weighting data from the secondary pages. This can improve runtime classification speed greatly.

Repeating Last Page

  • Contracts that contain signature pages being copied and distributed to involved parties and then signed and returned to be stored with the Contract document should use this feature.
  • Any document type that may or will have a duplicate last page.

Secondary Page Extractor

  • False-positive classification frequently happens on pages besides the first page of a document. In other words, middle pages and last pages, or secondary pages.
  • This property allows you to use an extractor to configure rules when to attach a secondary page to a particular Document Type whose first page has already been identified during ESP Auto Separation.
  • Use on multi-page Document Types using Unstructured Pagination. You can configure a Secondary Page Extractor on any To take advantage of this, configure a Positive Extractor in the Classification section and configure the Secondary Page Extractor to identify the second page. Cheating’s allowed in Grooper.

Examples

Separation will vary wildly between document types. Here are some real-world configurations using the new separation options. Some Classification Features are configured in the images. It’s common for Classification and Separation to be configured simultaneously. The explanations for each image consider classification and separation for runtime operations.

Separation and review 09.png

Positive Extractor
A Rule-Based Classification will occur and fallback on Features-Based Classification.

Pagination
It says this document can have any number of pages.

Prioritize EPI
This respects EPI if it is present, otherwise rely on EPI training from Features training.

Secondary Page Extractor
It determines if the page isn’t the first or last page of the document. If this extractor has a result, Separation appends this page to the page above it and understands the next page will be another secondary page, last page, or the start of a new document.

The Content Model has an EPI Extractor configured. not shown here

Separation and review 10.png

Allow Training
No lexical features (as set on the Content Model) are considered for classification or separation.

Positive Extractor
Separation will only occur when the Positive Extractor has a result. Rule-Based Classification only.

Pagination
Automatically appends a specified number of pages using training.

Repeating Last Page
In considering Structured Pagination, and this property set to True, repeating last pages will be appended.

Separation will separate using page-for-page trained samples only.
Copies of the last page, such as signed pages from multiple parties, may exist and will be appended.

Separation and review 11.png

Allow Training
Lexical features (as set on the Content Model) are considered for classification and separation.

Positive Extractor
A Rule-Based Classification will occur and fallback on Features-Based Classification.

Negative Extractor
A result from this extractor will exclude this Document Type as a classification option.

Pagination
This document can have any number of pages.

Training Scope
The features on page 1 will be the only features saved in training. If Rules-Based Classification fails, only the first page’s features of a trained sample are used. When separation occurs and detects page 1 of this document type, all proceeding pages will be appended until another recognized document type is identified.

The Content Model has an EPI Extractor configured and creates a hybrid of Rule-Based and Feature-Based Classification. not shown here

Separation and review 12.png

Allow Training
Lexical features (as set on the Content Model) are considered for classification and separation.

Positive Extractor
A Rule-Based Classification will occur and fallback on Features-Based Classification.

Negative Extractor
A result from this extractor will exclude this Document Type as a classification option.

Pagination
This document can have any number of pages.

Prioritize EPI
Respect EPI if it is present, otherwise rely on EPI training from Features training.

Secondary Page Extractor
- It determines if the page isn’t the first or last page of the document. If this extractor has a result, Separation appends this page to the page above it and understands the next page will be another secondary page, last page, or the start of a new document.

The Content Model has an EPI Extractor configured and creates a hybrid of Rule-Based and Feature-Based Classification. not shown here

Separation and review 13.png

Allow Training
Lexical features (as set on the Content Model) are considered for classification and separation.

Positive Extractor
A Rule-Based Classification will occur and fallback on Features-Based Classification.

Pagination
This document can have any number of pages.

Secondary Page Extractor
- It determines if the page isn’t the first or last page of the document. If this extractor has a result, Separation appends this page to the page above it and understands the next page will be another secondary page, last page, or the start of a new document.

The Content Model has an EPI Extractor configured and creates a hybrid of Rule-Based and Feature-Based Classification. not shown here

Separation Review

Use the Separation Review activity (in a Batch Process) any time a user will need to validate Separation and Classification is 100% accurate. The type of documents being processed (complexity, OCR result variances, etc.) can determine whether a user will need Separation Review.

Following is a (by no means exhaustive) list of industries whose documents have shown to frequently require Separation Review. These documents are considered unstructured document types due to complex and extremely variable language present on the document pages.

  • Oil & Gas
  • Legal
  • Healthcare


Documents that may optionally need Separation Review will depend on the priority of desired accuracy. These documents are considered structured document types because the pages contain predictable locations of repeated fields, tables, and sections used in both simple and complex classification techniques.

  • Accounts Recievable
  • Accounts Payable
  • Tax Documents
  • HR Documents
  • Healthcare

How To

Separation

Configuring Separation relies on understanding the different Separation Provider types and how to configure them.

Change in Value Separation

Please refer to this article on Change in Value Separation for information on this specific provider.

Control Sheet Separation

Please refer to this article on Control Sheet Separation for information on this specific provider.

EPI Separation

Please refer to this article on EPI Separation for information on this specific provider.

ESP Auto Separation

Please refer to this article on ESP Auto Separation for information on this specific provider.

Event-Based Separation

Please refer to this article on Event-Based Separation for information on this specific provider.

Multi Separator

Please refer to this article on Multi Separator for information on this specific provider.

Pattern-Based Separation

Please refer to this article on Pattern-Based Separation for information on this specific provider.

Undo Separation

Please refer to this article on Undo Separation for information on this specific provider.

Separation Review

Configuring Separation Review

Separation Review is a configuration of the Review activity of a Batch Process.

1. First, create a Batch Process.
2. Add a Batch Process Step.
3. Set the Activity Type property to Review.

Separation and review 31a.png

4. Click the Batch Views property then its ellipsis button.
5. In the Tab Page Setup Collection Editor window, click the Add button.

  • This will add an item to the list named <NO CONTROL TYPE>.

6. In the Batch View Type property drop-down, select Separation Viewer.

Separation and review 31b.png

7. Click the Control Settings property then its ellipsis button.
8. In the Separation Viewer Settings window click the drop-down arrow next to the Separation Settings property to expose further properties.
9. Click the Scope property, then it's corresponding drop-down arrow, and from the drop-down node tree, select a Content Type.

Separation and review 31c.png

Attachment Rules – Configure how certain document types should be attached to a regular document type.
Respect Original Page Numbers – If a PDF or mult-page TIF file is used to create individual pages, Grooper will keep track of the page’s original number and use that value to prevent the last page of one document from being attached to the first page of the next document.
Bind Only – If this is set to true, all pages will remain individual pages when Separation Review begins. If it is set to false, then Separation will attempt to classification and separation of pages into classified folders before Separation Review begins. As of this writing, setting this property to False seems to make Separation Review easier to complete.
Goto First Invalid Item – Separation Review will auto-select the first item that couldn’t be assigned a document type because that item’s confidence didn’t meet or exceed the threshold for confident classification set on the Content Model.
Flag Messages – Optional property to predefined acceptable messages to assign to items in Separation Review.

Interface Overview

Separation and review 14.png
Separation and review 15.png
The documents are highlighted with two alternating colors to easily distinguish how separation will occur without needing to check the name of the Document Type.

If an item has red text, this means the item has a classification flag. In this image, hovering over Page 9 would show a tooltip with the reason the item is flagged.

This image also shows pages because the Separate activity has the Bind Only property set to Truewhich will leave the pages as loose pages in the batch until choosing the Separate command at the end of the review process. If the Bind Only property is set to False, this image will show folders because the Separate activity proceeded to separate the pages into folders.

Separation and review 16.png
This is the Candidates list view showing a list of candidates and their respective confidences based on the current selection.

Appending and Prepending to Classified Groups of Documents

Right-clicking on Page 9 allows selection of the Prepend to Next object command. Ctrl+N is the shortcut key. Page 9 is now considered Page 1 of 4.
Separation and review 17.png
Separation and review 18.png
Right-clicking on Page 13 allows selection of Append to Previous object command. Ctrl+P is the shortcut key. Page 13 is now considered Page 5 of 5.
Separation and review 19.png
Separation and review 20.png

Multi-selecting Pages and Classifying Them

The keyboard should be the only control needed to work through Separation Review. When one or more pages are selected, start typing the name of the Document Type. This will filter the Candidates list view. Typing “paid” in this example, reduces the candidate list to five Document Types because their names have the word “paid” in them.

Separation and review 21a.png

When the desired Document Type is in the list, pressing the Tab button will cause the focus to change to the Candidates list view. Use the Up and Down buttons to highlight the correct document type.

Separation and review 22.png

Pressing Enter when the correct Document Type is selected will classify the pages to the highlighted (from the Candidate list view) document type.

Separation and review 23.png

Select Span

It is necessary at times to make a multi-selection to grab all the pages pertaining to a particular document. Using the mouse or keyboard to select a page, then shift or control select the remaining is possible, however unnecessary with the Select Span command. One need simply select a page of a document, and use the Select Span object command whose hotkey is Ctrl+S. With the Select Span object command used, all the pages of the document are selected. From here one could Ctrl+UP or Ctrl+DOWN hotkeys to move the entire document up or down in the list of documents.
Separation and review 24.png Separation and review 25.png

Completing Separation Review

Above the list of pages is the Separate button. It will only be enabled when all pages have a classification assigned.

-Starting Separation Review-
Separation and review 26.png

-When all pages are classified-
Separation and review 27.png

When all pages are classified and the Separate button is clicked, the list will change and show all pages separated into classified documents. Notice the Folder icons and Folder names.

Separation and review 28.png

Clicking the button next to the separate button will undo the separation and revert folders back to pages.

-All pages are separated into folers.-
Separation and review 29.png

Clicking the (blue arched arrow) Undo button will undo all manual classification performed during Separation Review and revert the batch of pages to their previous state when Separation Review started. Note that deleted pages will not be recovered. Deleted pages are permanently unrecoverable.

Property Details

Separation

Training Scope

This makes it possible to alter the separation training and logic if a Document Type has special attributes, pages, or layout that would normally confuse the ESP Separation logic.

  • Normal
    • This is the classic version of capturing training features for a document type.
  • FirstLast
    • Only trains the first and last page of a document type. See Use Cases for more information.
  • FirstOnly
    • Only trains the first page of a document type. See Use Cases for more information.

Repeating Last Page

Some Document Types require duplicate last pages. Normally, Separation would see a last page, complete the separation for a document, and leave duplicate last pages as loose pages. Enabling this property will reduce operator work in Separation Review by reducing the amount of page appending commands.

  • This property is only available when a Document Type’s Pagination is set to Structured.

Secondary Page Extractor

Any result created by this extractor identifies the page is not the first or last page of a document.

  • This property is only available when a Document Type’s Pagination is set to Unstructured.

Separation Review

Object Commands from the Context Menu

Separation and review 30.png

Append to Previous – Attaches selected item(s) to the preceding items above the selection.
Prepend to Next – Attaches selected item(s) to the proceeding items below the selection.
Delete – Deletes selected item(s).
Select Span – Shortcut command to highlight all items in the same color group.
Separate – Applies separation to items currently classified.
Unseparate – Removes separation to items currently in classified folders.
Move Up – Moves the selected item(s) up in the item order.
Move Down – Moves the selected items(s) down in the item order.
Flag Item – Assign a flag to an item for other workflow needs post-Separation Review.
Goto Next Loose Page – Auto-selects the next page not confidently assigned a document type.
Remove PDF Version – Removes an existing pdf file attached to the item.
Image – Provides a selection of image adjustment commands to apply to the current page.
Scan Once – Changes the current display image format.
Send To – Provides a selection of export commands to apply to the current page.
Properties – Displays metadata of a selected page.

Version Differences

Separation

Separation in Grooper 2.9 adds three new properties to Document Types.

  1. Training Scope
  2. Repeating Last Page
  3. Secondary Page Extractor

The Pagination property on a Document Type will determine if the new property will appear. See Property Details for more information.

Grooper 2.8 Grooper 2.9
Separation and review 01.png Separation and review 02.png
Grooper 2.8 Grooper 2.9
Separation and review 03.png Separation and review 04.png
Grooper 2.8 Grooper 2.9
Separation and review 05.png Separation and review 06.png
Grooper 2.8 Grooper 2.9
Separation and review 07.png Separation and review 08.png

Separation Review

Grooper 2.8 Classification Review

Grooper 2.8 and previous versions relied on Classification Review with its various controls to separate loose pages, merge selected pages into documents, and correct misclassified folders or classify unclassified folders. This user interface required a moderate-to-large effort to complete classification and separation, especially when reverting one or more folders to loose pages in order to combine the same pages into separate documents and classifying each newly created document.

Grooper 2.9 Separation Review

Starting with Grooper 2.9, Separation Review replaces Classification Review, which will also still be available for legacy support or other use cases. Don’t be concerned about this being a new module. Slap the name “Classification Review 2.0” on Separation Review because Separation Review really is a true upgrade and efficiently redesigned approach of Classification Review. Consider it a godsend!

The design of Separation Review invites the heavy use of the keyboard and keyboard shortcuts to make quick work of any items needing corrective action. Many quality-of-life improvements now exist and are quickly realized if one has previous experience using Classification Review.

Even though this is Separation Review, some classification tasks take place in this module. In the How-To section, items in the list are color coded. The list shows current classification of the items. This doesn’t necessarily mean they are currently separated into documents. The actual separation occurs later. During use of Separation Review, using the various context menu commands will assist and ensure pages are classified correctly and ready for proper separation.