การตัดคำ: ตัดให้เหลือรากศัพท์

The verb run shows up as running, runs, and ran. To you they share a meaning, but raw text counts each one separately, splitting the signal. 🌱

Stemming chops the endings off a word to reach a rough root called the stem. The goal is to make related forms collapse into one shared token.

A stemmer follows simple suffix rules instead of looking words up. It strips endings like ing, ed, and s, so it is fast but a bit crude.

NLTK ships the classic PorterStemmer. You create one instance, then call its stem method on any word you want reduced.

Feed in several forms of the same verb and they all reduce to the same stem. Now your counts can finally add up.

Stemming is blunt. The word studies becomes studi, which is not a real word, but it still matches study and studied as one stem.

Sometimes a stemmer cuts too far and merges unrelated words. This over-stemming can blur meanings, so always sanity-check the output on your text.

In practice you stem every token in a document. A comprehension applies the stemmer across the whole list in one clean line.

Porter is the default, but NLTK also offers the snappier SnowballStemmer, which supports many languages and fixes some Porter rough edges.

Stemmers expect lowercase input. Fold case before stemming, or Running and running may slip through as two different tokens again.

Stemming trades precision for speed. It shines in search and large pipelines where a rough but fast match beats slow, exact analysis. ⚡

Let's lock in how stemming behaves.

You learned stemming chops suffixes with simple rules to collapse word forms into one stem. It is fast and rough, sometimes yielding non-words. Nice work! 🎉