Part-of-speech tagging is the NLP task of assigning each word in a sentence its grammatical role. Noun, verb, adjective, adverb, pronoun and preposition labels get attached to the words. It looks simple but needs context, because the same word takes a different role depending on the sentence.
"Book" is a noun in "read the book" and a verb in "book the room", and nothing about the word itself settles that. English taggers lean heavily on word order and the surrounding words. Morphologically rich languages like Turkish need an extra step, splitting words into morphemes first and then choosing the right analysis from context.
Tags are rarely the product on their own, they feed other work. Dependency parsing, named entity recognition, intent analysis on search queries and stress placement in speech synthesis all consume this information.
A concrete case: a search engine receives the query "train times". Without knowing whether the first word is a noun or a verb the system cannot read the query correctly. Tagging settles it as a noun modifying the second, and the result set changes accordingly.
Today's language models do not run tagging as a separate step. Labelled data still gets used in classical NLP pipelines and in linguistic research, because the output stays inspectable and explainable.