Data split in Jasper

How the training data is split in Jasper

Let's say you uploaded 100 documents to your project and did the annotations and then you triggered training process.

Behind the scene Jasper engine will split the data into 2 parts. 90 of the uploaded documents will be used to train and build the model. 10 documents will be kept aside , once the model is build , Jasper will test the model against these 10 documents and compute the model performance metrics

Recall , Precision and F1 score is reported against how well the model performed against the 10 documents kept aside. Read More Build Your Model

🧙‍♂️ It's important to note that your model will not see or be trained on 10% of the total data you have uploaded to Jasper ,as the 10% is kept aside for validation