Multi-head attention is a term used in artificial intelligence and is primarily applied to big data, smart data and digital transformation. It is a special technique that utilises artificial intelligence models to better understand a large amount of data at the same time.
Think of Multi-Head Attention as a team of experts looking at different parts of a text at the same time and focussing on different aspects. Each expert draws their own conclusions and these are combined at the end to capture the big picture. For example, in a long customer dialogue, the system can recognise what the customer says at the beginning, but also how the conversation develops later - all information remains in view.
The major advantage of multi-head attention: machines and programs can analyse information much more precisely, recognise correlations better and provide more relevant answers. This helps, for example, to make chatbots more intelligent, improve automatic translations or analyse and understand large volumes of text in companies more quickly.
Deutsch 
English (UK)