lemmatisation

Lemmatization is the process of reducing a word to its base or dictionary form, the lemma. Unlike simple stemming, which may produce truncated or non-dictionary forms, lemmatization uses morphological analysis and, often, part-of-speech information to assign the most appropriate lemma for a given token in context.

In practice, a lemmatizer may rely on lexicons and inflection tables, rule-based morphological analyzers, or statistical models. A lemmatizer typically requires the surrounding context or a POS tag to select the correct lemma when a word has multiple possible lemmas (for example, "better" as an adjective may map to "good", while "better" as a verb maps to "improve").

Lemmatization is used in information retrieval to improve matching across word forms, in natural language processing pipelines for parsing and tagging, in question answering, machine translation, and text normalization.

Challenges include ambiguity, handling of proper nouns and multiword expressions, languages with rich morphology, and resource availability for less-resourced languages. Evaluation uses corpora annotated with lemmas or gold-standard morphological analyses.