Zero-Shot Text Classification

Why we started it?

Text Classification is a process of automatically assigning a label, or category, to a piece of text. This is usually done by using machine learning algorithms to analyze the text and learn from example data what words and phrases are associated with different categories.

Previously, separate models were needed for text classification into designated text categories. This means that a lot of effort had to be spent on labelling texts manually in order to create a dataset for the machine learning algorithm. However, with this NLP feature, a single model can be used for multiple text classification tasks, i.e. in ad-hock given classes/labels.

This is done by using a technique called transfer learning, which involves using a pre-trained model to initially learn the general structure of the data and then to find the most accurate class for the given text, among the provided ones.