Maximal Marginal Relevance (MMR) for information retrieval and summarization

E894311

diversity-based re-ranking method information retrieval technique ranking algorithm summarization technique

Maximal Marginal Relevance (MMR) is an information retrieval and summarization technique that selects results by jointly maximizing relevance to a query while minimizing redundancy among the chosen items.

Try in SPARQL Jump to: Surface forms Statements Referenced by

All labels observed (1)

Label	Occurrences
Maximal Marginal Relevance (MMR) for information retrieval and summarization canonical	1

Statements (52)

Predicate	Object
instanceOf	diversity-based re-ranking method ⓘ information retrieval technique ⓘ ranking algorithm ⓘ summarization technique ⓘ
abbreviation	MMR ⓘ
algorithmType	greedy selection algorithm ⓘ
alsoUsedFor	image retrieval diversification ⓘ video retrieval diversification ⓘ
appliedTo	document retrieval ⓘ extractive summarization ⓘ multi-document summarization ⓘ passage retrieval ⓘ query-focused summarization ⓘ recommendation systems ⓘ search result diversification ⓘ snippet selection ⓘ
assumes	access to pairwise similarity between items ⓘ
benefit	improves user-perceived diversity ⓘ increases coverage of different subtopics ⓘ reduces redundancy in result lists ⓘ
canUse	any similarity function satisfying basic properties ⓘ
category	diversity-aware ranking ⓘ redundancy reduction method ⓘ
coreIdea	penalize similarity to already selected items ⓘ trade off between query relevance and novelty ⓘ
field	information retrieval ⓘ natural language processing ⓘ text summarization ⓘ
goal	maximize relevance to a query ⓘ minimize redundancy among selected items ⓘ promote diversity in retrieved results ⓘ
hasParameter	lambda ⓘ
influenced	later diversification methods in IR ⓘ subtopic retrieval models ⓘ
introducedBy	Jade Goldstein NERFINISHED ⓘ Jaime G. Carbonell NERFINISHED ⓘ
introducedIn	paper "The Use of MMR, Diversity-Based Reranking for Reordering Documents and Producing Summaries" NERFINISHED ⓘ
lambdaControls	trade-off between relevance and diversity ⓘ
publicationYear	1998 ⓘ
publishedAt	SIGIR 1998 NERFINISHED ⓘ
relatedTo	coverage-based summarization ⓘ determinantal point processes NERFINISHED ⓘ novelty-based ranking ⓘ query-focused extractive summarization ⓘ
selectionCriterion	maximizes marginal gain in relevance minus redundancy ⓘ
selectionProcess	iteratively selects items ⓘ
typicalDomain	text documents ⓘ
typicalRepresentation	vector space model ⓘ
typicalSimilarityMeasure	cosine similarity GENERATED ⓘ
uses	linear combination of relevance and redundancy terms ⓘ similarity between candidate item and query ⓘ similarity between candidate item and selected items ⓘ

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Jaime Carbonell → notableWork → Maximal Marginal Relevance (MMR) for information retrieval and summarization ⓘ