Adaptive Document Retrieval for Deep Question Answering

TL;DR

This paper introduces an adaptive document retrieval method for deep question answering systems that dynamically determines the optimal number of documents to retrieve, improving performance across various datasets and corpus sizes.

Contribution

It proposes a novel adaptive retrieval model that learns to select the optimal number of candidate documents based on corpus size and query, addressing limitations of static retrieval strategies.

Findings

Outperforms state-of-the-art methods on multiple benchmarks

Effective across variable corpus sizes

Reduces noise-information trade-off in document retrieval

Abstract

State-of-the-art systems in deep question answering proceed as follows: (1) an initial document retrieval selects relevant documents, which (2) are then processed by a neural network in order to extract the final answer. Yet the exact interplay between both components is poorly understood, especially concerning the number of candidate documents that should be retrieved. We show that choosing a static number of documents -- as used in prior research -- suffers from a noise-information trade-off and yields suboptimal results. As a remedy, we propose an adaptive document retrieval model. This learns the optimal candidate number for document retrieval, conditional on the size of the corpus and the query. We report extensive experimental results showing that our adaptive approach outperforms state-of-the-art methods on multiple benchmark datasets, as well as in the context of corpora with variable sizes.