Linguistic annotation is the process of adding metadata to a corpus of text, which provides valuable information about linguistic features such as syntax, semantics, and pragmatics. It is essential for training and evaluating natural language processing models, enabling them to understand and generate human language more effectively.