R2-D2: A Modular Baseline for Open-Domain Question Answering

TL;DR

This paper introduces R2-D2, a four-stage open-domain question answering pipeline combining retrieval, reranking, extractive and generative reading, achieving state-of-the-art results on multiple datasets.

Contribution

The paper presents a modular four-stage QA system that effectively combines extractive and generative models, outperforming previous approaches.

Findings

Surpasses state-of-the-art on NaturalQuestions and TriviaQA datasets.

Combining extractive and generative readers improves accuracy by up to 5 exact matches.

A smaller extractive reader can match the performance of a larger generative reader.

Abstract

This work presents a novel four-stage open-domain QA pipeline R2-D2 (Rank twice, reaD twice). The pipeline is composed of a retriever, passage reranker, extractive reader, generative reader and a mechanism that aggregates the final prediction from all system's components. We demonstrate its strength across three open-domain QA datasets: NaturalQuestions, TriviaQA and EfficientQA, surpassing state-of-the-art on the first two. Our analysis demonstrates that: (i) combining extractive and generative reader yields absolute improvements up to 5 exact match and it is at least twice as effective as the posterior averaging ensemble of the same models with different parameters, (ii) the extractive reader with fewer parameters can match the performance of the generative reader on extractive QA datasets.