The term semi-supervised learning comes from the fields of artificial intelligence, big data, smart data and digital transformation. Semi-supervised learning means "partially supervised learning" and is a method that enables computers to learn independently from data.
In contrast to conventional supervised learning, where every piece of information (e.g. thousands of photos of cats and dogs, all clearly labelled) is labelled by a human, semi-supervised learning works with just a few labelled pieces of data and a large amount of unlabelled data. This saves a lot of time and money because not every single image or piece of information has to be checked by a human.
A simple example: Imagine a company wants to automatically sort its emails into "spam" and "non-spam". It only has 100 emails that are already marked as "spam" or "non-spam", but thousands are still unmarked. Semi-Supervised Learning helps to analyse these unmarked emails and improve the categorisation by recognising patterns.
This makes artificial intelligence more efficient and enables it to make valuable decisions more quickly, even if only little prepared data is available.
Deutsch 
English (UK)