[1] Create New Learning Model

Last modified by admin on 2024/01/11 18:17

Note: This feature is not available now.

Create new learning models to upload and train sample documents. After training is complete, send the learning models to production and use them to run on actual documents to extract data.

Staff can create a new learning model by following these below steps:

Step 1: Go to "Learning Models" tab

Step 2: Choose "Add Learning Model" button

Step 3: Input Model Name, Base Model (optional) and Description (optional)

Step 4: Create fields for models

If staff choose base model, this new model will get all the fields that the based model has and staff can add more fields by clicking button "Add field" or remove fields by unticking some fields.

If staff doesn't choose base model, staff will have to create form fields and table from scratch

Step 5: Click "Save" button

Note: Staff can not add, remove and configure fields after saving. So please check carefully before saving

Step 6: Click "Import" button

Step 7: Upload documents to train model and create tags for documents then click "OK" to upload

Step 8: Choose documents and label them sequently.

With the Form Fields, staff should label both "Label" and "Value" for each field if having enough data in documents. This helps the model will extract data more exactly.

Staff can choose "Label" or "Value" and then map them to suitable data in document

"Label" should be the value that has the same meaning to form field
"Value" should be the specific information of "Label"

With Table Fields, staff only need label "Label" and map it to the column name

Step 9: Click "Train Instance" button to start training model. After training models successfully, staff can use that model to extract data by creating new pipeline with document type is the model name that has been trained successfully

Note: Staff can bypass training some documents by using excluding documents when training function