Creating a SmartExtract Template

SmartExtract Templates allow you to automate the extraction of structured data from Documents. This article walks you through the steps to create a new SmartExtract Template, configure its fields, and prepare it for use.

 

The template is designed to allow you to give SmartExtract additional information to make value suggestions for each field nominated to be extracted.  

 

You do not have to enter additional information, SmartExtract has built in generic context.  However, the more contextual information SmartExtract has the more relevant the suggestions will be after reviewing a Document.

Navigate to: Configuration => Templates => SmartExtract Templates

Any existing templates will be listed in the SmartExtract Summary screen

Click [New]: The template screen will open for you to create your template

General

• Give your template a Name
• Optionally, update the 'Context for Type' field 
  • This option opens a pop up where you can enter a prompt to aid SmartExtract in initially identifying the correct Contract Type. 
• If you do not enter a prompt then the actual Type name will be used as context.
• For example, if you have a Contract Type of 'Order Form' then SmartExtract will search the document and if it finds text relating to 'Order Form' then it will suggest that this may be the correct Contract Type.

Fields

You can select the fields to be extracted based upon your accounts Contract Types.  

• Use the Type dropdown to choose the Contract Types this template applies to - this is a list of all your current Contract Types

• Once a Type is selected the screen will update and display all fields available within that Type layout

• You can tick and configure all the relevant fields for that Type that you wish SmartExtract to process
• Once that Type is completed you can select another Type and repeat
• The same template can be used to configure differing extraction fields based upon the Contract Type

The example below is a Customer Order Form template but the account has multiple 'Customer' Contract Types.

The same template can be used to configure all the 'Customer' Contract Types and then another template can be created for the 'Energy & Utilities' Contract Types if desired.

Field Context

For each field in the template you can:

• Ignore - (do not extract) leave the checkbox blank
• Extract - (include in extraction) tick the checkbox

If you choose to extract a field you can:

• Click Edit Context to open a pop up and enter information to to aid SmartExtract in identifying the correct information.
• Entering a Prompt
  • If you do not enter any information then the Field Title value will be used as context.
• Specifying a maximum number of characters of the returned suggestion (e.g. 30 for short text, 100 for long text, max 1000)
• Select Field Type 
  • The Field Type tells the SmartExtract what kind of data to expect. This ensures that the right extraction rules and validation logic. For example, if a field is set to "Date", the system will look for date formats in the document and ignore unrelated text.
  • Literal will extract the exact wording or phrase found in the document
  • Sentiment will analyse the surrounding context and provides a summarised or categorised response 

Default Values

If you have common fields then you can select them using the 'Default' Type 

Any Field selections and Field Context you have entered will automatically apply to other Types that have the same field.

Once all of your fields have been selected for each of the Types required [Save] the Template.

---

Note: For each Template, you must select at least one field for extraction.

---

The Summary screen will display your template and the Contract Types that are included in the template will identified in the 'Included Types' column.

If all Types apply to the template then 'All' will be displayed

Your template is now ready for use in SmartExtract

---

Tips for Success

• Use enhanced context for complex fields where deeper analysis is needed.
• Keep context clear and concise to improve extraction accuracy.

---