We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Rider: Reader-Guided Passage Reranking for Open-Domain Question Answering

Abstract: Current open-domain question answering systems often follow a Retriever-Reader architecture, where the retriever first retrieves relevant passages and the reader then reads the retrieved passages to form an answer. In this paper, we propose a simple and effective passage reranking method, named Reader-guIDEd Reranker (RIDER), which does not involve training and reranks the retrieved passages solely based on the top predictions of the reader before reranking. We show that RIDER, despite its simplicity, achieves 10 to 20 absolute gains in top-1 retrieval accuracy and 1 to 4 Exact Match (EM) gains without refining the retriever or reader. In addition, RIDER, without any training, outperforms state-of-the-art transformer-based supervised rerankers. Remarkably, RIDER achieves 48.3 EM on the Natural Questions dataset and 66.4 EM on the TriviaQA dataset when only 1,024 tokens (7.8 passages on average) are used as the reader input after passage reranking.
Comments: Minor format updates. TLDR: Reranking retrieved passages by reader predictions can achieve 10~20 gains in top-1 retrieval accuracy and 1~4 gains in Exact Match (EM) without any training
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
Cite as: arXiv:2101.00294 [cs.CL]
  (or arXiv:2101.00294v3 [cs.CL] for this version)

Submission history

From: Yuning Mao [view email]
[v1] Fri, 1 Jan 2021 18:54:19 GMT (36kb,D)
[v2] Sun, 30 May 2021 22:17:15 GMT (38kb,D)
[v3] Fri, 6 Aug 2021 20:25:40 GMT (38kb,D)

Link back to: arXiv, form interface, contact.