Lemmatization

Lemmatization is a text normalization technique used in natural language processing. The main difference from stemming is that lemmatization aims to remove endings to get to the root word, called the lemma, while considering context and vocabulary to ensure correctness.

Here are the key points about lemmatization:

It transforms words into their dictionary forms based on analysis of word morphology, prefix, suffix and part-of-speech.
For example, lemmatization would reduce "better" to its root form "good", whereas stemming may reduce it to "bet".
It uses a vocabulary database and morphologic analysis instead of just heuristic rules to get the canonical form of a word.
Popular lemmatization algorithms utilize WordNet and dictionaries to correctly convert verbs to their infinitive form, nouns to singular etc.
Lemmatization provides context and distinction between words with multiple meanings which stemming does not.
It is slower but more accurate than stemming and improves results for syntactic analysis, keyword extraction and other advanced NLP tasks.
Lemmatization works for highly inflected languages with complex morphology like German, Spanish etc. where stemming may be ineffective.

In summary, lemmatization considers context, dictionary definitions and morphology to transform words into root forms accurately, enabling better performance for search, IR and text analytics. It is more complex but resolves issues with ambiguity inherent in stemming.