When you click on an upload row, the panel with a detail view of the pipeline steps appears. You can close this panel via ‘ esc’. Via SHIFT-2 you can activate or deactivate the detail view. In this example we see input and output at upload level and for each document we see the step OCR, document classification, entity extraction and business rules. The icons can be green, red or orange for the known statuses.
We see the following results with the first document:
OCR was successful
Document classification has a certainty score of 100 (threshold 70%) and the document was recognised as a Dutch document of the type 'seat - dismissal - appointment'.
Ten entities were found and 16 were not.
In the second document we see that NS (Niels Sales) has performed manual intervention and that 4 business rules have been successfully validated and 3 have failed.
When page management is enabled you get an extra part. Here is an example of a project with page management and document classification, without entity extraction.
Here we see
The input was successful
The OCR was done at upload level and was successful.
Document creation (page management and document classification) had a security score of 84.74%, the threshold is 50%. Bart Van Mieghem performed a manual intervention in this step. He performed the following actions
23 pages were correct and have not changed (grey colour)
0 pages that were not recognized have been added to (green color)
0 pages are moved out of a document
0 pages have been moved to another document
6 pages have been removed from documents
A list of the final documents with their language.
DETAIL VIEW INPUT
You can click on each of the steps in the 'detail view pipeline' to see more details of that step. If you click on the 'input' step:
- If the files were uploaded via the UI - the file names (and link to the actual file) and show the user who uploaded the files via the UI.
- If the files have been uploaded via the API - show the body payload of your HTTP request to Metamaze with the corresponding response payload and response code.
Uploaded using the UI
Uploaded using the API
DETAIL VIEW PAGE MANAGEMENT
If you click on the page management step in the 'detail view pipeline' it will show the result of the page management model and what happened in the manual intervention.
Here you can see that several documents have been created from the files that have been uploaded. Each document in the list has a number of fields
An arrow to see the pages of the document
The identifier of the document
The document type
How many pages have not changed document type?
How many pages have been added to a new document?
How many pages have been moved from one document to another?
How many pages have been moved to another document?
How many pages have been deleted from a document
The same icons and colors are used to indicate pages that have been changed as you can see in the image above.
The top arrow allows you to open all documents at once to see if they all close at once.
You can always click on a page to see a preview of the page.
DETAIL VIEW DOCUMENT
When you click on a document, you get a detailed view of the entity extraction and business rules validation. This view consists of two parts:
The entity extraction component shows all entities that may or may not be extracted by the model or manually in the document. It consists of three parts, all separated by a horizontal line:
The entities not found in the document (grey colour)
The entities found by the model with a certainty score greater than the threshold.
The entities approved in manual intervention (green tick), removed (red cross) or added manually (plus icon).
These lists of entities have different columns:
The pane icon shows whether a business rule has been defined for this entity and whether the validation failed (red color) or was successful (green color). Clicking this icon will open the business line in the business line table so you can see what the result of the condition was.
The entity type
The certainty score that the model returned and the set threshold between round brackets. If the score was lower than the threshold it will be shown in red.
The entity value found in the document
The converted value to the requested format
Clicking on an entity will show it in an additional panel with the document.
At the bottom there is a list of all business rules. The list has following columns:
A plus icon to view the result of the condition.
The name of the business rule
Has the rule been successfully validated or not
You can click on the plus icon to view the business rule in detail
SEND DOCUMENTS INDIVIDUALLY OR IN BATCH TO TRAINING
In the last column of the 'documents' overview you will see checkboxes and/or arrows.
Checking a box of one or more documents and clicking on 'sent to training' will allow you to sent documents manually to training even though they were automatically processed. Documents that were validated in human validation will be automatically sent to training.
An arrow indicates this document was already sent to training, either manually or automatically. Clicking on the arrow will link you to the document in the training module.
You can find these documents also manually by filtering on status 'to verify' or creating a task for documents with status 'to verify'. They will need to be verified before they are promoted to 'golden' training data.