---
title: "Hybrid retrieval (PT)"
source: "https://systems-analysis.info/int/Hybrid_retrieval_(PT)"
wiki: "systems-analysis.info/int"
article: "Hybrid_retrieval_(PT)"
language: "pt"
categories:
  - "Category:Large language models"
  - "Category:Portuguese"
  - "Category:Prompt engineering"
revision_id: 3094
wiki_created_at: 2026-09-06T23:15:36Z
wiki_modified_at: 2026-09-06T23:15:36Z
downloaded_at: 2026-09-07T22:55:04Z
---

# Hybrid retrieval (PT)

**Recuperação Híbrida (Hybrid Retrieval)** — é uma classe de métodos de recuperação de informação que combina sinais lexicais (sparse) e semânticos (dense/late-interaction) para aumentar o recall e a precisão dos resultados. Esquemas híbridos unem as vantagens da correspondência exata de termos (BM25/TF-IDF) e da proximidade vetorial (bi-encoders, modelos multivetoriais de interação tardia), além de utilizar métodos de fusão de classificações robustos a pontuações de diferentes escalas (como Reciprocal Rank Fusion, CombSUM/CombMNZ) e re-ranking com cross-encoders.<sup>[\[1\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-1)[\[2\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-2)[\[3\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-3)</sup>

## Definição e Motivação

*Recuperação híbrida* é uma busca paralela ou em cascata por dois (ou mais) canais de sinais independentes, seguida por fusão e/ou re-ranking. As motivações típicas incluem: (i) superar a "lacuna de vocabulário" (sinônimos, reformulações), (ii) robustez a erros de digitação/morfologia, (iii) extração de códigos/identificadores específicos (onde modelos sparse são fortes), e (iv) generalização para novos domínios/idiomas (onde modelos dense fornecem generalização semântica).<sup>[\[4\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-4)[\[5\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-5)[\[6\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-6)</sup>

## Componentes da Busca Híbrida

### Lexical (sparse)

- **Modelos clássicos.** TF-IDF e BM25/BM25F são métodos de base padrão que utilizam índices invertidos; o BM25 é fundamentado no framework probabilístico PRF e é amplamente usado no primeiro estágio de ranqueamento.<sup>[\[7\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-7)</sup>
- **Modelos sparse treináveis.**
  - **SPLADE / SPLADE++/v3.** Um modelo neuro-sparse que aprende a expandir e ponderar termos através de uma cabeça MLM com regularização de esparsidade; demonstra resultados fortes e boa capacidade de generalização (BEIR).<sup>[\[8\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-8)[\[9\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-9)[\[10\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-10)</sup>
  - **uniCOIL/COIL.** Listas invertidas contextualizadas e sua versão simplificada, *uniCOIL*; são compatíveis com índices invertidos clássicos.<sup>[\[11\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-11)</sup>

### Semântico (dense/late-interaction)

- **Bi-encoder (single-vector).** A consulta e o documento são codificados por modelos vetoriais, e a similaridade é calculada por produto escalar (dot-product)/MIPS. Exemplos: DPR,<sup>[\[12\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-12)</sup> ANCE,<sup>[\[13\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-13)</sup> Contriever,<sup>[\[14\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-14)</sup> GTR,<sup>[\[15\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-15)</sup> E5.<sup>[\[16\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-16)</sup>
- **Late-interaction (multi-vector).** Modelam correspondências em nível de token durante a interação "tardia": ColBERT/ColBERTv2; o trade-off é uma maior precisão com um índice/latência maior, o que é mitigado por otimizações de engenharia (PLAID, WARP).<sup>[\[17\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-17)[\[18\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-18)[\[19\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-19)</sup>

## Esquemas de Hibridização e Fusão de Ranks

- **Busca paralela e união de candidatos.** Listas de candidatos (sparse e dense) são obtidas independentemente com suas pontuações internas; em seguida, as classificações são fundidas.<sup>[\[20\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-20)</sup>
- **RRF (Reciprocal Rank Fusion).** Uma técnica robusta a pontuações de classificações incomparáveis, que soma os ranks recíprocos:

${RRF}(d) = \sum\limits_{i = 1}^{m}\frac{1}{k + {rank}_{i}(d)}$, onde geralmente $k \approx 60$.<sup>[\[21\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-21)</sup> É suportada em motores de busca industriais (Elasticsearch/OpenSearch) como um recuperador/processador integrado.<sup>[\[22\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-22)[\[23\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-23)</sup>

- **CombSUM/CombMNZ e outros.** Funções clássicas de "soma de pontuações" (com normalização, se necessário).<sup>[\[24\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-24)[\[25\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-25)[\[26\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-26)</sup>
- **Combinação linear ponderada.**

$S(d) = \alpha \cdot S_{\text{sparse}}(d) + (1 - \alpha) \cdot S_{\text{dense}}(d)$, $\alpha \in \lbrack 0,1\rbrack$. A escolha de $\alpha$ pode ser fixa ou treinável (por coleção/por consulta).<sup>[\[27\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-27)</sup>

- **Normalização de pontuações.** Para CombSUM/CombMNZ, métodos como min-max, z-score, etc., são frequentemente usados para alinhar as escalas;<sup>[\[28\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-28)</sup> alternativamente, o RRF baseia-se apenas nos ranks.
- **Ponderação dinâmica/adaptativa.** Roteamento de consultas (query routing), características da consulta e modelos LTR (learning-to-rank) para selecionar/ponderar os canais; trabalhos recentes mostram que uma simples combinação ponderada e treinada frequentemente supera o RRF e é pouco sensível à normalização.<sup>[\[29\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-29)</sup>

## Re-ranking e Pipelines Multi-estágio

Sistemas híbridos são geralmente construídos como *retrieval → fusion → rerank*. Para o re-ranking, são utilizados:

- **Cross-encoders (BERT/T5).** Os mais precisos, porém custosos: MonoBERT/MonoT5 para reordenar os top N candidatos.<sup>[\[30\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-30)[\[31\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-31)</sup>
- **Late-interaction como re-ranker.** A família ColBERT também pode atuar como re-ranker; aceleradores modernos (PLAID, WARP) reduzem a latência sem perda de qualidade.<sup>[\[32\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-32)[\[33\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-33)</sup>

O trade-off *qualidade ↔ latência/custo* é especialmente importante em RAG e com SLAs rigorosos (ver latências de cauda p95/p99).<sup>[\[34\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-34)</sup>

## Avaliação em Benchmarks

- **BEIR.** Um conjunto unificado de coleções/tarefas heterogêneas para avaliação zero-shot/out-of-domain de recuperadores (ex., TREC‑COVID, NFCorpus, NQ, HotpotQA, FiQA‑2018, DBPedia-entity, ArguAna, Webis‑Touché‑2020, FEVER/Climate-FEVER, Scidocs, SciFact, CQADupStack, etc.).<sup>[\[35\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-35)</sup>
- **TREC Deep Learning / MS MARCO.** Recursos clássicos para treinar/avaliar recuperadores e re-rankers em cenários de big data.<sup>[\[36\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-36)[\[37\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-37)[\[38\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-38)</sup>
- **Métricas de qualidade.** nDCG@k, Recall@k, MRR; para desempenho — latência p50/p95/p99, QPS; para operação — memória/custo (CPU/GPU, índice).<sup>[\[39\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-39)[\[40\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-40)</sup>
- **Ablações.** Recomenda-se isolar a contribuição de cada canal/peso e a sensibilidade aos parâmetros $k$ no RRF e $\alpha$ na combinação ponderada; avaliar a robustez a reformulações e a mudanças de distribuição (OOD).<sup>[\[41\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-41)[\[42\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-42)</sup>

## Aspectos de Engenharia e Práticas de Produção

- **Índices e ANN.** FAISS (Flat/HNSW/IVF‑PQ), HNSW, ScaNN para MIPS/similaridade de cosseno.<sup>[\[43\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-43)[\[44\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-44)[\[45\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-45)</sup>
- **Stack de IR.** Lucene/Anserini/Pyserini para pipelines sparse, dense e híbridos; reprodutibilidade "com dois cliques" no BEIR.<sup>[\[46\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-46)[\[47\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-47)</sup>
- **Bancos de dados vetoriais e motores de busca.** Qdrant, Weaviate, pgvector/PostgreSQL, Vespa, Elasticsearch/OpenSearch possuem modos nativos de busca híbrida (BM25F+vetor) e/ou RRF/combinação linear.<sup>[\[48\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-48)[\[49\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-49)[\[50\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-50)[\[51\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-51)[\[52\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-52)</sup>
- **Padrão RAG.** Arquitetura: **retrieval → fusion → rerank → contexto do LLM** com limitação de tokens e rastreamento de fontes.<sup>[\[53\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-53)</sup>
- **Atualização de índices, desduplicação, tokenização.** É importante alinhar a tokenização entre o BM25 и o vetorizador; calibração de pontuações (normalização/escalonamento) antes da combinação.<sup>[\[54\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-54)</sup>

## Limitações e Questões Abertas

- **Generalização e multilinguismo.** Modelos dense (GTR/E5) melhoram a generalização, mas são sensíveis ao domínio/idioma; modelos sparse (SPLADE) são frequentemente mais robustos em cenários OOD.<sup>[\[55\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-55)[\[56\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-56)</sup>
- **Integração com LLMs e alucinações.** A recuperação híbrida reduz omissões e ruído nos contextos de RAG, mas não elimina completamente as alucinações; são necessários re-rankers rigorosos e filtragem de fontes.<sup>[\[57\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-57)</sup>
- **Custo e privacidade.** Armazenamento de índices multi-vetoriais, compressão, criptografia e stack on-premise; avaliação do TCO.
- **Tendências.** HyDE/doc2query/PRF como expansão de documentos/consultas;<sup>[\[58\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-58)[\[59\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-59)</sup> aprendizado da combinação (per-query $\alpha$), modelos late-interaction mais eficientes (PLAID/WARP), documentos longos e índices multi-vetoriais.<sup>[\[60\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-60)[\[61\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-61)</sup>

## Tabela Comparativa de Métodos

Dados de 2025‑09‑10 (exemplo na coleção BEIR *trec‑covid*; nDCG@10 / Recall@100):<sup>[\[62\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-62)</sup>

| Método                      | Tipo (sparse/dense/híbrido) | Ideia/Modelo                                     | Esquema de fusão                   | Re-ranker            | nDCG@10 / R@100                                                               | Latência (rel.) | Fontes                                                                                                                                                                                                                                                |
|-----------------------------|-----------------------------|--------------------------------------------------|------------------------------------|----------------------|-------------------------------------------------------------------------------|-----------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| BM25                        | sparse                      | Correspondência exata de termos (PRF/BM25)       | —                                  | —                    | 0.595 / 0.109                                                                 | muito baixa     | <sup>[\[63\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-63)[\[64\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-64)</sup>                                                                               |
| SPLADE++ (ED)               | sparse (learned)            | Expansão/pesos esparsos de termos                | —                                  | —                    | 0.727 / 0.128                                                                 | baixa-média     | <sup>[\[65\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-65)[\[66\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-66)</sup>                                                                               |
| Contriever (MS MARCO FT)    | dense                       | Bi-encoder de aprendizado contrastivo            | —                                  | —                    | 0.596 / 0.091                                                                 | média           | <sup>[\[67\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-67)[\[68\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-68)</sup>                                                                               |
| BGE‑base‑en‑v1.5            | dense                       | Embedder universal forte                         | —                                  | —                    | 0.781 / 0.141                                                                 | média           | <sup>[\[69\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-69)</sup>                                                                                                                                                             |
| Cohere embed‑english‑v3.0   | dense                       | Modelo de embedding de texto de nível industrial | —                                  | —                    | 0.818 / 0.159                                                                 | média           | <sup>[\[70\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-70)</sup>                                                                                                                                                             |
| BM25 + dense (ex: BM25+BGE) | hybrid                      | Recuperação paralela + fusão de listas           | RRF (k≈60) ou combinação ponderada | opc.: MonoT5/ColBERT | (varia conforme a implementação; geralmente \> que o melhor canal individual) | média           | <sup>[\[71\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-71)[\[72\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-72)[\[73\]](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_note-73)</sup> |

Comparação de métodos em *trec‑covid*

Nota: a última linha ilustra o esquema; os números exatos dependem da escolha do embedder, da normalização e dos parâmetros de fusão (consulte as fontes e os scripts reprodutíveis do Pyserini).

## Ligações externas

- Pyserini / Anserini: <a href="https://github.com/castorini/pyserini" class="external text" rel="nofollow">github.com/castorini/pyserini</a> • <a href="https://github.com/castorini/anserini" class="external text" rel="nofollow">github.com/castorini/anserini</a>
- FAISS: <a href="https://arxiv.org/abs/1702.08734" class="external text" rel="nofollow">arXiv:1702.08734</a>
- Weaviate (Hybrid search): <a href="https://docs.weaviate.io/weaviate/search/hybrid" class="external text" rel="nofollow">docs.weaviate.io/weaviate/search/hybrid</a>
- pgvector: <a href="https://github.com/pgvector/pgvector" class="external text" rel="nofollow">github.com/pgvector/pgvector</a>
- Vespa (Hybrid search tutorial): <a href="https://docs.vespa.ai/en/tutorials/hybrid-search.html" class="external text" rel="nofollow">docs.vespa.ai/en/tutorials/hybrid-search.html</a>

## Literatura

- Manning, C.D., Raghavan, P., Schütze, H. (2008). *Introduction to Information Retrieval*. Cambridge University Press. ISBN 978‑0521865715.
- Robertson, S., Zaragoza, H. (2009). *The Probabilistic Relevance Framework: BM25 and Beyond*. Foundations and Trends in Information Retrieval 3(4):333–389. DOI:10.1561/1500000019.
- Lin, J. et al. (2021). *Pyserini: A Python Toolkit for Reproducible IR*. SIGIR.
- Järvelin, K., Kekäläinen, J. (2002). *Cumulated Gain‑Based Evaluation of IR Techniques*. Information Retrieval 6:241–256. DOI:10.1023/A:1016043826386.
- Dean, J., Barroso, L.A. (2013). *The Tail at Scale*. CACM 56(2):74–80. DOI:10.1145/2408776.2408794.

## Notas

1.  <span id="cite_note-1">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-1) Robertson, S., Zaragoza, H. (2009). *The Probabilistic Relevance Framework: BM25 and Beyond*. Foundations and Trends in Information Retrieval, 3(4), 333–389. DOI:10.1561/1500000019.</span>
2.  <span id="cite_note-2">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-2) Cormack, G.V., Clarke, C.L.A., Büttcher, S. (2009). *Reciprocal Rank Fusion Outperforms Condorcet and Individual Rank Learning Methods*. SIGIR 2009, 758–759. <a href="https://cormack.uwaterloo.ca/cormacksigir09-rrf.pdf" class="external text" rel="nofollow">PDF</a>.</span>
3.  <span id="cite_note-3">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-3) Bruch, S., Gai, S., Ingber, A. (2023). *An Analysis of Fusion Functions for Hybrid Retrieval*. ACM TOIS 42(1):1–35. DOI:10.1145/3596512 • arXiv:2210.11934.</span>
4.  <span id="cite_note-4">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-4) Manning, C.D., Raghavan, P., Schütze, H. (2008). *Introduction to Information Retrieval*. Cambridge University Press. ISBN 978‑0521865715 (ver capítulos sobre TF‑IDF, avaliação e o problema de *vocabulary mismatch*).</span>
5.  <span id="cite_note-5">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-5) Izacard, G. et al. (2022). *Unsupervised Dense Information Retrieval with Contrastive Learning (Contriever)*. TACL 10:1089–1108. arXiv:2112.09118.</span>
6.  <span id="cite_note-6">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-6) Wang, L. et al. (2022/2024). *Text Embeddings by Weakly‑Supervised Contrastive Pre‑training (E5)*. arXiv:2212.03533.</span>
7.  <span id="cite_note-7">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-7) Robertson, S., Zaragoza, H. (2009). *The Probabilistic Relevance Framework: BM25 and Beyond*. DOI:10.1561/1500000019.</span>
8.  <span id="cite_note-8">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-8) Formal, T., Piwowarski, B., Clinchant, S. (2021). *SPLADE: Sparse Lexical and Expansion Model for First Stage Ranking*. arXiv:2107.05720.</span>
9.  <span id="cite_note-9">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-9) Formal, T. et al. (2022). *Making Sparse Neural IR Models More Effective*. Findings of EMNLP. arXiv:2205.04733.</span>
10. <span id="cite_note-10">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-10) Formal, T. et al. (2024). *SPLADE‑v3: New baselines for SPLADE*. arXiv:2403.06789.</span>
11. <span id="cite_note-11">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-11) Lin, J., Ma, X. (2021). *A Few Brief Notes on DeepImpact, COIL, and uniCOIL*. arXiv:2106.14807.</span>
12. <span id="cite_note-12">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-12) Karpukhin, V. et al. (2020). *Dense Passage Retrieval for Open‑Domain QA*. EMNLP. arXiv:2004.04906.</span>
13. <span id="cite_note-13">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-13) Xiong, L. et al. (2021). *Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval (ANCE)*. ICLR. arXiv:2007.00808.</span>
14. <span id="cite_note-14">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-14) Izacard, G. et al. (2022). TACL. arXiv:2112.09118.</span>
15. <span id="cite_note-15">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-15) Ni, J. et al. (2021/2022). *Large Dual Encoders Are Generalizable Retrievers (GTR)*. EMNLP. arXiv:2112.07899.</span>
16. <span id="cite_note-16">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-16) Wang, L. et al. (2022/2024). arXiv:2212.03533.</span>
17. <span id="cite_note-17">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-17) Khattab, O., Zaharia, M. (2020). *ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT*. SIGIR. arXiv:2004.12832.</span>
18. <span id="cite_note-18">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-18) Santhanam, K. et al. (2022). *ColBERTv2 & PLAID*. NAACL/ArXiv. arXiv:2112.01488; arXiv:2205.09707.</span>
19. <span id="cite_note-19">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-19) Scheerer, J.L. et al. (2025). *WARP: An Efficient Engine for Multi‑Vector Retrieval*. arXiv:2501.17788.</span>
20. <span id="cite_note-20">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-20) Lin, J. et al. (2021). *Pyserini: A Python Toolkit for Reproducible IR with Sparse and Dense Representations*. SIGIR. <a href="https://cs.uwaterloo.ca/~jimmylin/publications/Lin_etal_SIGIR2021_Pyserini.pdf" class="external text" rel="nofollow">PDF</a>.</span>
21. <span id="cite_note-21">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-21) Cormack, G.V., Clarke, C.L.A., Büttcher, S. (2009). SIGIR. <a href="https://cormack.uwaterloo.ca/cormacksigir09-rrf.pdf" class="external text" rel="nofollow">PDF</a>.</span>
22. <span id="cite_note-22">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-22) Elastic Docs. *Reciprocal Rank Fusion*. (acesso em 2025‑09‑10). <a href="https://www.elastic.co/docs/reference/elasticsearch/rest-apis/reciprocal-rank-fusion" class="external text" rel="nofollow">elastic.co/docs/.../reciprocal-rank-fusion</a>.</span>
23. <span id="cite_note-23">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-23) OpenSearch Docs. *Score ranker processor (RRF)*. (acesso em 2025‑09‑10). <a href="https://docs.opensearch.org/3.0/search-plugins/search-pipelines/score-ranker-processor/" class="external text" rel="nofollow">docs.opensearch.org/.../score-ranker-processor/</a>.</span>
24. <span id="cite_note-24">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-24) Fox, E.A., Shaw, J.A. (1994). *Combination of Multiple Searches*. TREC‑2, NIST SP 500‑215, 243–252. <a href="https://www.khoury.northeastern.edu/home/jaa/CSG339.06F/resources/fox94combination.pdf" class="external text" rel="nofollow">PDF</a>.</span>
25. <span id="cite_note-25">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-25) Lee, J.H. (1997). *Analyses of Multiple Evidence Combination*. SIGIR, 267–276. DOI:10.1145/258525.258587.</span>
26. <span id="cite_note-26">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-26) Hsu, D.F., Taksa, I. (2005). *Comparing Rank and Score Combination Methods for Data Fusion in IR*. (Tech. report). <a href="https://ccc.inaoep.mx/~villasen/bib/Hsu-FusionInIR07.pdf" class="external text" rel="nofollow">PDF</a>.</span>
27. <span id="cite_note-27">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-27) Bruch, S., Gai, S., Ingber, A. (2023). TOIS. DOI:10.1145/3596512.</span>
28. <span id="cite_note-28">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-28) Hsu, D.F., Taksa, I. (2005). ver acima.</span>
29. <span id="cite_note-29">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-29) Bruch, S., Gai, S., Ingber, A. (2023). TOIS. DOI:10.1145/3596512.</span>
30. <span id="cite_note-30">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-30) Nogueira, R., Cho, K. (2019). *Passage Re‑ranking with BERT*. arXiv:1901.04085.</span>
31. <span id="cite_note-31">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-31) Nogueira, R., Jiang, Z., Lin, J. (2020). *Document Ranking with a Pretrained Sequence‑to‑Sequence Model (MonoT5)*. Findings of EMNLP. arXiv:2003.06713.</span>
32. <span id="cite_note-32">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-32) Santhanam, K. et al. (2022). arXiv:2205.09707.</span>
33. <span id="cite_note-33">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-33) Scheerer, J.L. et al. (2025). arXiv:2501.17788.</span>
34. <span id="cite_note-34">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-34) Dean, J., Barroso, L.A. (2013). *The Tail at Scale*. CACM 56(2):74–80. DOI:10.1145/2408776.2408794.</span>
35. <span id="cite_note-35">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-35) Thakur, N. et al. (2021). *BEIR: A Heterogeneous Benchmark for Zero‑shot Evaluation of IR Models*. NeurIPS Datasets & Benchmarks. arXiv:2104.08663.</span>
36. <span id="cite_note-36">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-36) Craswell, N. et al. (2020). *Overview of the TREC 2019 Deep Learning Track*. arXiv:2003.07820.</span>
37. <span id="cite_note-37">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-37) Craswell, N. et al. (2021). *Overview of the TREC 2020 Deep Learning Track*. arXiv:2102.07662.</span>
38. <span id="cite_note-38">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-38) Bajaj, P. et al. (2016). *MS MARCO: A Human Generated MAchine Reading COmprehension Dataset*. arXiv:1611.09268.</span>
39. <span id="cite_note-39">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-39) Järvelin, K., Kekäläinen, J. (2002). *Cumulated Gain‑Based Evaluation of IR Techniques*. Information Retrieval 6:241–256. DOI:10.1023/A:1016043826386.</span>
40. <span id="cite_note-40">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-40) Dean, J., Barroso, L.A. (2013). CACM. DOI:10.1145/2408776.2408794.</span>
41. <span id="cite_note-41">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-41) Bruch, S. et al. (2023). DOI:10.1145/3596512.</span>
42. <span id="cite_note-42">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-42) Ni, J. et al. (2021/2022). arXiv:2112.07899.</span>
43. <span id="cite_note-43">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-43) Johnson, J., Douze, M., Jégou, H. (2017). *Billion‑scale Similarity Search with GPUs (FAISS)*. arXiv:1702.08734.</span>
44. <span id="cite_note-44">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-44) Malkov, Y., Yashunin, D. (2020). *HNSW*. IEEE TPAMI 42(4):824–836. DOI:10.1109/TPAMI.2018.2889473.</span>
45. <span id="cite_note-45">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-45) Guo, R. et al. (2020). *ScaNN: Efficient Vector Similarity Search at Scale*. arXiv:1908.10396.</span>
46. <span id="cite_note-46">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-46) Yang, P., Fang, H., Lin, J. (2018). *Anserini: Reproducible IR Research with Lucene*. JDIQ 10(4):1–20. DOI:10.1145/3239571.</span>
47. <span id="cite_note-47">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-47) Lin, J. et al. (2021). SIGIR. <a href="https://cs.uwaterloo.ca/~jimmylin/publications/Lin_etal_SIGIR2021_Pyserini.pdf" class="external text" rel="nofollow">PDF</a>.</span>
48. <span id="cite_note-48">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-48) Qdrant Docs. *Hybrid queries (RRF, DBSF)*. (acesso em 2025‑09‑10). <a href="https://qdrant.tech/documentation/concepts/hybrid-queries/" class="external text" rel="nofollow">qdrant.tech/.../hybrid-queries/</a>.</span>
49. <span id="cite_note-49">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-49) Weaviate Docs. *Hybrid search*. (acesso em 2025‑09‑10). <a href="https://docs.weaviate.io/weaviate/search/hybrid" class="external text" rel="nofollow">docs.weaviate.io/weaviate/search/hybrid</a>.</span>
50. <span id="cite_note-50">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-50) pgvector GitHub. (acesso em 2025‑09‑10). <a href="https://github.com/pgvector/pgvector" class="external text" rel="nofollow">github.com/pgvector/pgvector</a>.</span>
51. <span id="cite_note-51">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-51) Vespa Docs. *Hybrid Text Search Tutorial*. (acesso em 2025‑09‑10). <a href="https://docs.vespa.ai/en/tutorials/hybrid-search.html" class="external text" rel="nofollow">docs.vespa.ai/.../hybrid-search.html</a>.</span>
52. <span id="cite_note-52">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-52) Elastic Docs. *Reciprocal Rank Fusion*. (acesso em 2025‑09‑10). <a href="https://www.elastic.co/docs/reference/elasticsearch/rest-apis/reciprocal-rank-fusion" class="external text" rel="nofollow">elastic.co/docs/.../rrf</a>.</span>
53. <span id="cite_note-53">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-53) Lewis, P. et al. (2020). *Retrieval‑Augmented Generation for Knowledge‑Intensive NLP Tasks*. NeurIPS. arXiv:2005.11401.</span>
54. <span id="cite_note-54">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-54) Hsu, D.F., Taksa, I. (2005). ver acima.</span>
55. <span id="cite_note-55">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-55) Ni, J. et al. (2021/2022). arXiv:2112.07899.</span>
56. <span id="cite_note-56">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-56) Formal, T. et al. (2021, 2022, 2024). arXiv:2107.05720; 2205.04733; 2403.06789.</span>
57. <span id="cite_note-57">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-57) Lewis, P. et al. (2020). arXiv:2005.11401.</span>
58. <span id="cite_note-58">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-58) Gao, L. et al. (2023). *Precise Zero‑Shot Dense Retrieval without Relevance Labels (HyDE)*. ACL. arXiv:2212.10496.</span>
59. <span id="cite_note-59">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-59) Nogueira, R. et al. (2019). *Document Expansion by Query Prediction*. arXiv:1904.08375; *docTTTTTquery*. <a href="https://cs.uwaterloo.ca/~jimmylin/publications/Nogueira_Lin_2019_docTTTTTquery-v2.pdf" class="external text" rel="nofollow">PDF</a>.</span>
60. <span id="cite_note-60">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-60) Santhanam, K. et al. (2022). arXiv:2205.09707.</span>
61. <span id="cite_note-61">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-61) Scheerer, J.L. et al. (2025). arXiv:2501.17788.</span>
62. <span id="cite_note-62">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-62) Pyserini *BEIR Regressions* (acesso em 2025‑09‑10): resultados em *trec‑covid* para BM25/SPLADE/Contriever/BGE/Cohere. <a href="https://castorini.github.io/pyserini/2cr/beir.html" class="external text" rel="nofollow">castorini.github.io/pyserini/2cr/beir.html</a>.</span>
63. <span id="cite_note-63">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-63) Robertson, S., Zaragoza, H. (2009). DOI:10.1561/1500000019.</span>
64. <span id="cite_note-64">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-64) Pyserini BEIR. Ver link acima.</span>
65. <span id="cite_note-65">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-65) Formal, T. et al. (2021, 2022). arXiv:2107.05720; 2205.04733.</span>
66. <span id="cite_note-66">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-66) Pyserini BEIR.</span>
67. <span id="cite_note-67">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-67) Izacard, G. et al. (2022). arXiv:2112.09118.</span>
68. <span id="cite_note-68">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-68) Pyserini BEIR.</span>
69. <span id="cite_note-69">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-69) Pyserini BEIR.</span>
70. <span id="cite_note-70">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-70) Pyserini BEIR.</span>
71. <span id="cite_note-71">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-71) Cormack et al. (2009). SIGIR. RRF.</span>
72. <span id="cite_note-72">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-72) Bruch et al. (2023). TOIS.</span>
73. <span id="cite_note-73">[↑](https://systems-analysis.info/int/Hybrid_retrieval_(PT)#cite_ref-73) Elastic/OpenSearch RRF Docs.</span>