Staple AI also supports the ability for users to create their document types and train the AI to capture data points of their choosing. This is done using models. Models are like custom-made templates for extracting data from documents.

Step 1: Initialising Custom models

Creating a custom model is simple. First, navigate to the “Models” section on the leftmost section of the page. From here the user selects the “New Model” button. This leads to a popup asking for details to create a new model.

The options for the model are:

1. Model name: Enter the name to identify the particular model
2. Image type: Choose the processing type for the image. A dropdown has the following options, No image processing (used when the user does not want to process the image), A4, Slip, Card (these are used for specific image type) and Rotate only (this option processes the image in two planes, while others process the image as it is)
3. Language: This prompts the user to select whether the image has to be translated or not.
4. Members added: Here, the user invites people with whom this model is to be shared.

The fields of model name, image type and language are compulsory as they help in processing the entered documents.

‍

Step 2: Creating a model:

Once a model is initialized, creating its template is the next task. Doing so marks all the fields a user wants to extract from the document.

Add a document that contains all the desired fields. The document will look like as shown below:

Figure 24: Creating a model

Notice the fields list is empty. This is for the user to enter their desired fields for scanning.

This is done by clicking on the “New field” button. This will list 4 mandatory sections to be filled.

1. Name of field: This marks the name of the field. This will then be the name of the field in the scanning area - and in the exported data (csv/json/xml) as the key.
2. Keyword/ Phrase: From the document, which keyword you as a person reading the document look for to understand the desired data is linked here. User has to link this from already existing phrases in the document.
3. Data area: This links to the data from the document. This is the actual data that should be extracted from the document (as opposed to the keyword/phrase). More than one areas can be linked
4. Type: A dropdown is shown from which the user selects the type of data to be searched for. For example - should the extracted data ALWAYS be a date? If so the date option should be chosen.
   

‍

Step 3: Adding tables

Adding tables to be scanned is just as easy. Hover over to the “Smart Tables” section to the right of the “Fields” section. Following it, add a new table using the button “New Smart table”. This creates a table on the document and two categories that will identify the table. The user can adjust the table to align with the one on the document. Then, the categories on the left need to be filled up:

1. Name of the table: This field takes a text that identifies the table to be scanned.
2. Title of table: This creates a link to the document that the AI should look for to identify the table.

While processing the table, a header of a particular column can only be updated if it is mentioned in the category list of the header. The “New Category” button adds a new category to the existing list of headers.

Users can add multiple fields or tables. An example is shown below.

‍

Step 4: Using Custom models

Once a model is created, it may be used in exactly the same way as a default model in the scanning section. While creating or updating a queue, the user must select the custom model option under the “Model type”. See Creating a Queue for more information. After that, select the model to be used under the “Document type”. This will apply the model to the documents in the queue.

Step 4.1:  Updating existing models:

If there  are changes to be made in the model, a user

can do so by clicking on the three dots alongside the model name in the list of models. Choosing the “Edit model” option will lead the user to a popup where the changes can be made.

Step 4.2: Deleting Custom models:

To delete a model select the option “Remove model” that appears after clicking the same three dots mentioned above.

‍