Stopsõnade

Stopsõnade (stop words) are common function words that occur very frequently in a language and typically carry little lexical meaning on their own. In text processing and information retrieval they are often removed from texts before indexing or querying to reduce dimensionality and computational cost, and to improve the signal-to-noise ratio.

Because languages differ, stop word lists are language-specific and domain-specific. A list for Estonian will include common conjunctions, pronouns, prepositions and auxiliary verbs that appear across texts, such as ja (and), on (is), see (this), et (that), kuid (but), või (or), ka (also), kui (if).

Removal is not universal: some tasks benefit from retaining stop words to preserve syntax, discourse structure, or sentiment cues. In English and Estonian alike, stemming or lemmatization is often used before or after stop word removal to handle inflected forms.

In Estonian, the high level of inflection means that many stop words have multiple morphological variants; effective lists may rely on lemmas or include common inflected forms.

Applications include search engines, document indexing, and text mining, where removing stop words reduces index size and speeds up queries, though the exact approach depends on language, domain, and task.

Debate exists about the value of stop words for specific tasks; modern NLP often uses lighter or task-specific lists rather than a fixed universal set.