Answering Complex Open-Domain Questions with Multi-Hop Dense Retrieval

By Wenhan Xiong, Xiang Lorraine Li, Srini Iyer, Jingfei Du, Patrick Lewis, William Yang Wang, Yashar Mehdad, Wen-tau Yih, Sebastian Riedel, Douwe Kiela and 1 more. 2020

246 citations recorded by the source corpus.

We propose a simple and efficient multi-hop dense retrieval approach for answering complex open-domain questions, which achieves state-of-the-art performance on two multi-hop datasets, HotpotQA and multi-evidence FEVER. Contrary to previous work, our method does not require access to any corpus-specific information, such as inter-document hyperlinks or human-annotated entity markers, and can be applied to any unstructured text corpus. Our system also yields a much better efficiency-accuracy trade-off, matching the best published accuracy on HotpotQA while being 10 times faster at inference time.

Subfields: Artificial Intelligence, Information Systems, Information Retrieval & Search, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

In Information Retrieval & Search

In Natural Language Processing

Citations in the corpus

Internet-augmented language models through few-shot prompting for open-domain question answering (2022, neutral): “We report accuracy for the classiﬁcation tasks (left ﬁgure) and exact match for the generation tasks (right ﬁgure).”
Retrieval Augmentation Reduces Hallucination in Conversation (2021, methodology): “FiD has been applied with great success to open-domain QA tasks primarily with BM25 retrievers or neural retrievers pre-trained on QA datasets (Izacard and Grave, 2020; Xiong et al., 2021).”
Adaptive Semiparametric Language Models (2021, methodology): “…there are several recent retrieval-augmented methods that have been used for question answering (de Masson d’Autume et al., 2019; Guu et al., 2020; Xiong et al., 2021; Kassner and Schutze, 2020), controllable generation (Xu et al., 2020), machine translation (Bapna and Firat, 2019; Khandelwal et…”
Adaptive Information Seeking for Open-Domain Question Answering (2021, methodology): “For example, some works employ a single retrieval function multiple times (Das et al., 2019a; Qi et al., 2019; Xiong et al., 2021), and the other works use a pre-deﬁned sequence of retrieval functions (Asai et al., 2020; Dhingra et al., 2020).”
Boosting Search Engines with Interactive Agents (2021, neutral): “Xiong et al. (2021) successfully use relevance feedback by jointly encoding the question and the text of its retrieved results for multi-hop QA.”
UniK-QA: Unified Representations of Structured and Unstructured Knowledge for Open-Domain Question Answering (2020, background): “For multi-hop question answering, there is recent work (Xiong et al., 2021b) successfully extending dense retrieval to the multi-hop setting (Yang et al., 2018; Welbl et al., 2018), which could naturally be applied within our framework.”
The Fact Extraction and VERification Over Unstructured and Structured information (FEVEROUS) Shared Task (2021, methodology): “Bust a move! applied a three stage-process to the task consisting of multi-hop dense passage-retrieval (Xiong et al., 2021) trained on FEVEROUS data, followed by BM25 ﬁltering to ensure that sentences containing named entities mentioned in the claim are not ranked below those sentences semantically…”
Exploiting Reasoning Chains for Multi-hop Science Question Answering (2021, methodology): “Comparison Methods: For fair comparison, we compare CGR with recently published methods that use similar power of pretrained models, including five textual knowledge based methods: AristoRoBERTa (AllenAI, 2019), KF-SIR (Banerjee and Baral, 2020), FreeLB (Zhu et al., 2020), DPR (Karpukhin et al., 202”
Iteratively Prompt Pre-trained Language Models for Chain of Thought (2022, methodology): “We are also inspired by methods that iteratively retrieve knowledge from explicit knowledge sources for multi-step reasoning, such as (Xiong et al., 2021; Qi et al., 2019; Khattab et al., 2021; Mo et al., 2022b).”
Retrieve, Rerank, Read, then Iterate: Answering Open-Domain Questions of Arbitrary Complexity from Text (2020, background): “Firstly, some previous work assumes the access to a well-tuned parameterized retrieval system that helps navigate the large amount of text (Das et al., 2019; Lee et al., 2019; Asai et al., 2020; Guu et al., 2020; Khattab et al., 2020; Xiong et al., 2020).”
Dense Text Retrieval Based on Pretrained Language Models: A Survey (2022, methodology): “More recently, without using hyperlinks, multi-hop dense retrieval [331], [332] has been proposed to enhance the performance of complex QA tasks [57].”
Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions (2022, methodology): “Xiong et al. (2021) extends the widely-used Dense Passage Retrieval (DPR) (Karpukhin et al., 2020) to multihop setting, which has since been improved by Khattab et al. (2021).”
A Metaverse: Taxonomy, Components, Applications, and Open Challenges (2022, unknown): “Xiong et al. [291] proposed a simple and efﬁcient multi-hop dense search approach to answer complex open-domain questions, achieving state-of-the-art performance in two multi-hop data sets, HotpotQA and multi-evidence FEVER.”
Knowledge Graph Prompting for Multi-Document Question Answering (2023, methodology): “To overcome this problem, we follow the training strategy of MDR (Xiong et al. 2020) and train a sentence encoder by predicting the subsequent supporting facts based on previously supporting facts, thereby endowing the encoder with reasoning capability.”
Case-Based Abductive Natural Language Inference (2020, background): “such as open-domain question answering (Xiong et al., 2021; Fang et al., 2020; Yang et al., 2018), increasing attention is being dedicated to the abstractive setting, where the models are required to compose long chains of facts expressing abstract commonsense and scientific knowledge (Jhamtani and ”
Baleen: Robust Multi-Hop Reasoning at Scale via Condensed Retrieval (2021, background): “, 2020) or single-vector dot-product (Xiong et al., 2020) retrievers, whose capacity to model a single open-domain question is limited (Khattab et al.”
Retrieving and Reading: A Comprehensive Survey on Open-domain Question Answering (2021, methodology): “For those with Dense Retriever, most prior systems including Multi-step Reasoner [29], MUPPET [35] and MDR [85] use MIPS retrieval to obtain the most semantically similar documents given a representation of the question; Path Retriever [88] develops a Recurrent Neural Network (RNN) retrieval to lear”