To avoid data labelling, we can utilise zero-shot learning that aims to perform modelling using less amount of labelled data. When this learning comes to text classification, we call the whole process zero-shot text classification.