Passage Retrieval 관련 연구

PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval

• 요약
  • Document-to-Document 검색을 위한 PARM(Paragraph Aggregation Retrieval Model)과 PARM에 적합한 순위 집계 함수 VRRF(Vector-based aggregation with Reciprocal Rank Fusion) 제안
• 연도: 2022
• 논문 url: [2201.01614] PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval
• github: GitHub - sophiaalthammer/parm: This repository contains the code for the paper 'PARM: Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval' published at ECIR'22.

Pre-training Methods in Information Retrieval

• 요약
• 연도: 2021
• 논문: https://arxiv.org/abs/2111.13853
• github

Sparsifying Sparse Representations for Passage Retrieval by Top- Masking

• 요약
• 연도: 2021
• 논문: [2112.09628] Sparsifying Sparse Representations for Passage Retrieval by Top-$k$ Masking
• github

Encoder Adaptation of Dense Passage Retrieval for Open-Domain Question Answering

• 요약
• 연도: 2021
• 논문: [2110.01599] Encoder Adaptation of Dense Passage Retrieval for Open-Domain Question Answering
• github

EntQA: Entity Linking as Question Answering

• 요약
• 연도: 2021
• 논문: [2110.02369] EntQA: Entity Linking as Question Answering
• github

Boosted Dense Retriever

• 요약
• 연도: 2021
• 논문: [2112.07771] Boosted Dense Retriever
• github

Learning Discrete Representations via Constrained Clustering for Effective and Efficient Dense Retrieval

• 요약
  • RepCONC 제안
• 연도
• 논문: [2110.05789] Learning Discrete Representations via Constrained Clustering for Effective and Efficient Dense Retrieval
• github

Baleen: Robust multi-hop reasoning at scale via condensed retrieval

• 요약
• 연도: 2021
• 논문: Baleen: Robust Multi-Hop Reasoning at Scale via Condensed Retrieval
• github: GitHub - stanford-futuredata/Baleen: Baleen: Robust Multi-Hop Reasoning at Scale via Condensed Retrieval (NeurIPS'21)

Efficient Passage Retrieval with Hashing for Open-domain Question Answering

• 요약
• 연도: 2021
• 논문: [2106.00882] Efficient Passage Retrieval with Hashing for Open-domain Question Answering
• github: GitHub - studio-ousia/bpr: Binary Passage Retriever (BPR) - an efficient passage retriever for open-domain question answering

Rethink training of BERT rerankers in multi-stage retrieval pipeline

• 요약
  • loss 함수로 LCE(Localized Contrastive Estimation) 제안
• 연도: 2021
• 논문 url: [2101.08751] Rethink Training of BERT Rerankers in Multi-Stage Retrieval Pipeline
• github:

COIL: Revisit Exact Lexical Match in Information Retrieval with Contextualized Inverted List

• 요약
  • query와 passage 사이의 스코어 계산하는 COIL 제안
  • query encoder와 passage encoder의 [CLS] 토큰 벡터를 곱한 값과 모든 토큰의 벡터 중에 가장 큰 벡터를 합하여 스코어를 계산함
• 연도: 2021
• 논문 url: https://arxiv.org/abs/2104.07186
• github: GitHub - luyug/COIL: NAACL2021 - COIL Contextualized Lexical Retriever

A Few Brief Notes on DeepImpact, COIL, and a Conceptual Framework for Information Retrieval Techniques

• 요약
  • COIL에서 토큰 차원을 1로 변경한 uniCOIL 제안
• 연도: 2021
• 논문 url: https://arxiv.org/abs/2106.14807
• github:

In-batch negatives for knowledge distillation with tightly-coupled teachers for dense retrieval

• 요약
  • TCT-ColBERT
• 연도: 2021
• 논문 url: In-Batch Negatives for Knowledge Distillation with Tightly-Coupled Teachers for Dense Retrieval - ACL Anthology
• github:

Efficiently teaching an effective dense retriever with balanced topic aware sampling

• 요약
  • TAS-B
• 연도: 2021
• 논문 url: https://arxiv.org/abs/2104.06967
• github:

Complementing lexical retrieval with semantic residual embedding

• 요약
• 연도: 2021
• 논문 url: [2004.13969] Complementing Lexical Retrieval with Semantic Residual Embedding
• github

Learning Passage Impacts for Inverted Indexes

• 요약
  • DeepImpact
• 연도: 2021
• 논문: [2104.12016] Learning Passage Impacts for Inverted Indexes
• github:

Approximate nearest neighbor negative contrastive learning for dense text retrieval

• 요약
  • ANCE
• 연도: 2021
• 논문: [2007.00808] Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval
• github:

Complement Lexical Retrieval Model with Semantic Residual Embeddings

• 요약
• 연도: 2021
• 논문:
• github

RocketQA: An optimized training approach to dense passage retrieval for open-domain question answering

• 요약
  • RocketQA
• 연도: 2021
• 논문: RocketQA: An Optimized Training Approach to Dense Passage Retrieval for Open-Domain Question Answering - ACL Anthology
• github

ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction

• 요약
  • ColBERTv2
• 연도: 2021
• 논문: [2112.01488] ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction
• github

Dense passage retrieval for open-domain question answering

• 요약
  • BERT 기반의 biencoder 구조 모델로 question과 passage 사이의 스코어를 계산할 수 있는 DPR(Dense Passage Retrieval) 제안
• 연도: 2020
• 논문: https://arxiv.org/abs/2004.04906
• github: GitHub - facebookresearch/DPR: Dense Passage Retriever - is a set of tools and models for open domain Q&A task.

Sparse, dense, and attentional representations for text retrieval

• 요약
  • Me-BERT
• 연도: 2020
• 논문:
• github

ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT

• 요약
  • COlBERT
• 연도: 2020
• 논문: https://arxiv.org/abs/2004.12832
• github

Improving efficient neural ranking models with cross-architecture knowledge distillation

• 요약
  • DistilBERT
• 연도: 2020
• 논문: [2010.02666] Improving Efficient Neural Ranking Models with Cross-Architecture Knowledge Distillation
• github

Efficient document re-ranking for transformers by precomputing term representations.

• 요약
• 연도: 2020
• 논문: https://arxiv.org/abs/2004.14255
• github

Context-Aware Sentence/Passage Term Importance Estimation For First Stage Retrieval

• 요약
  • DeepCT
• 연도: 2019
• 논문: [1910.10687] Context-Aware Sentence/Passage Term Importance Estimation For First Stage Retrieval
• github