IBM Document Processing

IBM Document Processing classifies and extracts information from your business documents more quickly, easily and accurately. This document describes how to use IBM Document Processing skills on IBM watsonx Orchestrate.

Using IBM Document Processing skills

To use the IBM Document processing skills, you must first add them to your skill set.

  1. On the watsonx Orchestrate chat page, click Add skills from the catalog.
  2. Search and click the IBM Document Processing app tile.
  3. You can add the skill in two ways:
    • Click Add skill + in the skill tile or select the skill tile and click the Add skill button.
  4. Next, open the watsonx Orchestrate chat.
  5. On the chat, start the IBM Document Processing skill.
  6. Click Connect app.
  7. You are prompted to enter the following connection details:
    • Bearer Token: You can specify any dummy data as a placeholder. watsonx Orchestrate internally manages the connection to IBM Document Processing.
  8. Click Connect app.

To know the available skills that you can use, see the list of prebuilt apps, skills and skill flows.

Technical parameters

The following table contains the technical parameters required by some skills of this app and instructions to retrieve them.

Skill name: Classify a document

Classifies a document into bill_of_lading, contract, invoice, purchase_order, utility_bill, or tax_form.

Input

Field Type Definition
File content File The document to be classified. The supported formats are:
  • .pdf
  • .docx
  • .pptx
  • .jpg
  • .png
  • .tiff
Language String The language in which the document was written. The supported languages are:
  • en: For documents in English. By default, English is selected.
  • fr: For documents in French.

Output

Field Type Definition
Class String The document classification, it can be one of:
  • bill_of_lading
  • contract
  • Invoice
  • purchase_order
  • utility_bill
  • tax_form
Confidence Float The confidence percentage of the document classification.
Classifications Array List of all possible document classifications. The document classifications class_name and confidence are displayed for each instance in that list.

Skill name: Extract invoice fields (KVPs)

Extracts information from invoice documents.

Input

Field Type Definition
File content File The invoice document from which information is to be extracted. The supported formats are:
  • .pdf
  • .docx
  • .pptx
  • .jpg
  • .png
  • .tiff
Language String The language in which the document was written. The supported languages are:
  • en: For documents in English. By default, English is selected.
  • fr: For documents in French.

Output

Field Type Format Definition
Bank account number String The bank account number.
Bank name String Name of the bank.
Bill-to address String The address used to send the invoice.
Bill-to name String Name of the person who receives the invoice.
Invoice date String Date The invoice date.
Invoice number Number The invoice number.
Invoice total Number The total invoice amount.
Payment due date String Date The invoice payment due date.
Payment terms String The invoice payment terms.
Purchase order number String The purchase order number.
Ship-to address String The address to send the items listed on the invoice.
Ship-to name String The address to send the items listed on the invoice.
Shipping amount Number The total cost to send the items.
Subtotal Number The sum of the items price listed in the invoice.
Tax amount Number The tax amount.
Tax rate Number The tax rate.
Tax type String The tax type.
Vendor address String The vendor address.
Vendor name String The vendor name.

Skill name: Extract utility bill fields (KVPs)

Extracts information from utility bills.

Input

Field Type Definition
File content File The content can be a .pdf, MS Office document (.doc, .docx, .ppt, .pptx), or an image file (.jpg, .png, .tiff).
Language String The language in which the document was written. The supported languages are:
  • en: For documents in English. By default, English is selected.
  • fr: For documents in French.

Output

Field Type Format Definition
account_number String Account number details.
amount_due Number Amount due details.
client_number String Client number details.
company_name String Company name details.
company_address String Company address details.
customer_name String Customer name details.
customer_address String Customer address details.
due_date String Date Due date details.
payment_received Number Payment received details.
previous_balance Number Previous balance details.
service_address String Service address details.
statement_date String Date Statement date details.

Skill name: Extract personal identifiable information (PII)

Detects potential personal identifiable information (PII) mentioned in a raw text fragment.

Input

Field Type Definition
Text String The raw text to extract potential PII entities.
Language String The language in which the document was written. The supported languages are:
  • en: For documents in English. By default, English is selected.
  • fr: For documents in French.

Output

Field Type Definition
Entities Array A list of potentials PII mentions. Each item of the list contains text, type, location, begin, and end.
Text String The text fragment that is detected as potential PII.
Type String The type of PII detected. For more information, see PII Entities documentation.
Location Object The offsets of the PII entity text. It shows the begin and end offset of the text fragment.
begin Integer The start offset of the text fragment that is detected as potential PII. The first character of the extracted text is at offset 0.
end Integer The end offset of the text fragment that is detected as potential PII. The last character of the mention is at offset - 1.

Skill name: Extract text

Extracts the text from a document.

Input

Field Type Definition
File content File The document to be extracted in text. The supported formats are:
  • .pdf
  • .docx
  • .pptx
  • .jpg
  • .png
  • .tiff
Language String The language in which the document was written. The supported languages are:
  • en: For documents in English. By default, English is selected.
  • fr: For documents in French.

Output

Field Type Definition
Extracted text String The text extracted from the document.
Info: Review the known issues for Prebuilt apps.

Parent topic:

→ Prebuilt apps