part-of-speech tagging

Part-of-speech tagging is the chore of labeling every word in a sentence with its grammatical role — noun, verb, adjective, and so on — the very thing you learned to diagram in school. Take "Time flies like an arrow": a tagger marks "Time" as a noun, "flies" as a verb, "like" as a preposition, and so forth. It is one of the oldest, most basic jobs in language processing, and it sits near the bottom of many bigger pipelines.

The reason it is harder than it sounds is that the same word can play different parts depending on its company. "Book" is a noun in "read a book" but a verb in "book a flight." "Flies" is a verb in "time flies" but a noun in "fruit flies are pests." A good tagger can't just memorize a dictionary; it has to weigh the surrounding words. Modern taggers learn this from large collections of text where humans have already marked the correct tags, then generalize to new sentences — a classic case of sequence labeling, where each word gets its own label in context.

Why bother? Knowing the parts of speech helps machines untangle who-did-what-to-whom, pick out key phrases, disambiguate meanings, and feed cleaner signals into translation, search, and text-to-speech (so a reader knows whether "read" is present or past tense). On well-edited English, taggers now top 97% accuracy — but that last few percent hides the genuinely ambiguous cases, and performance can sag on informal text, new slang, or under-studied languages.