fullword

Fullword is a term used in computing and linguistics to denote a complete lexical word as a unit, rather than a morpheme, stem, or subword fragment. It is not a standardized technical term across all fields, but it is used in documentation and discussions about text processing to contrast full words with subword units.

In search and indexing, fullword matching requires that the query match a complete word boundary in the text. This avoids false positives from partial matches. For example, a fullword search for "cat" would not find "category," while a substring search would. Implementations often rely on tokenization to split text into words and on boundary rules to determine matches.

In natural language processing, "fullword" can refer to tokens that correspond to dictionary entries, as opposed to affixes or clitics treated as separate tokens, or to the base units produced by tokenization. Some NLP pipelines use subword units (byte-pair encoding, character n-grams) for model input, which are not full words; others maintain fullword tokens for certain tasks, typically after lemmatization.

In lexicography and corpora, corpora may annotate words as "fullword" tokens and mark inflected variants as separate occurrences of the same lemma. Limitations include handling hyphenated compounds and languages with rich morphology.

See also: tokenization, stemming, lemmatization, full-text search, exact match, substring search.