HI @Anonymous
We have a doc update in progress trying to give our customers a better explanation about how to prepare data for category classification. It is going to be published soon, but let me summarize some good tips here.
Before you build a category classification model - AI Builder | Microsoft Docs
1. each tag needs to have 10 distinct text examples
2. each tag needs to have a minimum of 10 distinct text samples where it isn't used. So if you have a file of 20 records, but only 1 tag, it's not going to train.
3. A table needs to have at least 2 tags, each one must have 10 text samples.
4. You can define up to 200 distinct tags. Each tag is a category that will be identified and extracted from the given text.
5. Each sample of text data must have fewer than 5,000 characters.
So a minimum possible set here is 20 records, 1 tag per record, and 10 distinct examples for one tag, 10 distinct examples for the rest tag.
If you still encounter same issue, please email us at aihelpen@microsoft.com. And please help create an idea in the ideas forum to raise the request of better error messaging and fix suggestions, it will help us prioritize the work item.
Thanks
Norah