Training models

To train a model in Classify, you'll need a dataset. A dataset is a Zip file containing folders with examples.
Let's suppose you want to train an email classifier. These emails are split into 3 categories: claims, contract update, product question.

To create the training dataset, you just have to create a folder "email_classif" and 3 folders names "claims", "contract update" and "product question" copy at least 50 example emails in each, so 50 real claims emails in "claims", 50 contract update tin "contract update" and 50 product question emails in "product question", like for instance the tree below:

email_classif
claims
 angry customer.eml
 are you sleeping or what.eml
 ...(48 more)
contract update
 address update.eml
 name change after marriage.eml
 ...(52 more)
product question
 what is the interest rate.eml
 how to change the beneficiary.eml
 ...(61 more)

Just zip this structure and send it to Classify. A few moment later, the model is ready to be called with the model's API.


Did this page help you?