It's actually quite technical, but essentially, the transformer architecture helps each part of the sentence “talk” to all the other parts at the same time. This way, each part can understand what the whole sentence is about and what it means.

Here is the paper that imo changed the world 6 years ago and is the reason for the current state of AI.

https://arxiv.org/abs/1706.03762

If it goes over your head (it did for me), ask bing or chatgpt to summarise it for you. It helped me get my head around this stuff, as I'm in no way an expert nor do I study this field.