I'm trying to train a multi-label Custom Classification model on Amazon Comprehend. I uploaded a CSV where in the first column, I have the labels (e.g., PRODUCT | DELIVERY) and right after, the comment (text) that is related to the topics.
Example CSV row:
DELIVERY | SERVICE | PRODUCT , wanted to get more of the same product but it was out of stock. I was sad because I couldn't buy more and couldn't even pick it up at the desired store (I had to choose the store where the product was available for delivery).
I'm receiving the error: failed caused by Insufficient data. Required: At least 10 examples for each label. Consider adding more training data., exit code: 1
I have a total of 8 topics, all with well over 10 examples, but they are mixed as in the example above, since most comments address more than one topic.
I want to understand if, by having more than one topic, a "new topic" is generated, where the algorithm will understand it as one thing, or if it could be another error.