queryuitbreiding

Queryuitbreiding, in information retrieval, refers to methods for enhancing a user’s original query by adding additional terms. The aim is to reduce vocabulary mismatch between the query and the documents, thereby improving retrieval performance. By expanding a query, systems seek to increase recall while attempting to preserve or improve precision.

Common approaches fall into several categories. Lexical expansion uses dictionaries, thesauri, or domain ontologies to add synonyms or related concepts. Statistical or data-driven expansion relies on the content of retrieved documents or large corpora: pseudo-relevance feedback selects terms from top-ranked results to broaden the query, while co-occurrence or association measures identify terms that frequently appear with query terms. Language-model approaches adjust the probabilities of terms in the query, and embedding-based methods use semantic similarity in vector spaces to propose related words or phrases. Cross-language expansion translates or maps query terms into other languages to retrieve multilingual content.

A typical expansion workflow involves an initial retrieval with the original query, extraction and weighting of candidate expansion terms, and a second retrieval using the expanded query. Some systems enable user input to select or prune expansions to limit drift. Benefits include improved recall and better handling of synonyms, misspellings, and morphological variants. Risks include query drift, where expansions shift the focus away from the user’s intent, and increased computational cost or noise.

Query expansion remains a central technique in search engines, digital libraries, and question-answering systems. Modern approaches increasingly combine lexical, statistical, and neural methods to balance coverage with precision.