# Candidate Generation and Ranking **Domain:** Information Retrieval / Natural Language Processing / Search **Doc Type:** Canonical Mechanism Node **Maturity:** Developed **Related:** [[wiki/Information Retrieval|Information Retrieval]], [[wiki/Approximate Phonological Matching|Approximate Phonological Matching]], [[wiki/Constraint Satisfaction|Constraint Satisfaction]], [[wiki/Vector Retrieval|Vector Retrieval]], [[wiki/Language Model|Language Model]] --- ## Definition **Candidate Generation and Ranking separates the production of plausible options from the more expensive task of ordering them by relevance or quality.** Candidate generation protects recall; ranking supplies discrimination. ## Language-Engine Context A rhyme engine can use phonological indexes to locate a broad neighborhood, then rank candidates by stress, terminal alignment and correspondence length. A semantic engine can traverse associations to generate possible concepts, then rank them by relation to the author's primary subject and recent context. The distinction also supports transparency. The system can report why an item entered the candidate set and which criteria moved it upward. This is stronger than returning a single unexplained answer. ## Contemporary Bridge Modern retrieval systems often combine lexical indexes, graph traversal and [[wiki/Vector Retrieval|vector retrieval]] before reranking. A language model can contribute candidates but should not erase whether a candidate was retrieved, inferred or generated. ## Key Insight **Finding possibilities and judging possibilities are different operations with different failure modes.** ## See Also [[wiki/Information Retrieval|Information Retrieval]], [[wiki/Approximate Phonological Matching|Approximate Phonological Matching]], [[wiki/Constraint Satisfaction|Constraint Satisfaction]], [[wiki/Vector Retrieval|Vector Retrieval]], [[wiki/Semantic-First Constrained Generation|Semantic-First Constrained Generation]], [[wiki/Provenance-Sensitive Multiplicity|Provenance-Sensitive Multiplicity]] ## Sources / Provenance - Primary project account: [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]]. - Information-retrieval and search architecture. - Project reconstruction of phrase-level rhyme and associative traversal.