The general idea is to classify texts that the model has never seen before, or in other words, allow us to assign an appropriate label to a piece of text without having received any training examples before. This approach is also called zero-shot learning.
...
There are multiple advantages, including:
✅ No need to be fine-tunedfor training data. Fast way to classify text without taking up your time for additional annotation of examples, solving the problem of “cold start”.
✅ Unlimited possibilities for labels. Your model will be able to detect any label you might be interested in finding in your data, such as emotions, different topics, sentiment, personality traits and others.
...
Classes or also called labels, between which to discriminate (classify) the text, e.g.
Status colour Red title NEGATIVE Status colour Green title POSITIVE Patterns, also known as label descriptors, allow the semantic matching between the analysed text and the different labels, e.g.
This text is {}
Input: The product has no issues but the packaging causes so much extra to squirt out and you can't stop it. For how expensive it is it's such a waste.
Result:
| Status | ||||
|---|---|---|---|---|
|
...