It's a way of measuring the similarity between objects, such as two long-form articles, by converting the content into numbers and then plotting them on a diagram. Closer objects are more-similar than objects that are further apart.

Vector embeddings are lists of those calculated numbers.