@inproceedings{wu-etal-2020-dont-read,
title = "Don{'}t Read Too Much Into It: Adaptive Computation for Open-Domain Question Answering",
author = "Wu, Yuxiang and
Minervini, Pasquale and
Stenetorp, Pontus and
Riedel, Sebastian",
editor = "Moosavi, Nafise Sadat and
Fan, Angela and
Shwartz, Vered and
Glava{\v{s}}, Goran and
Joty, Shafiq and
Wang, Alex and
Wolf, Thomas",
booktitle = "Proceedings of SustaiNLP: Workshop on Simple and Efficient Natural Language Processing",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2020.sustainlp-1.9",
doi = "10.18653/v1/2020.sustainlp-1.9",
pages = "63--72",
abstract = "Most approaches to Open-Domain Question Answering consist of a light-weight retriever that selects a set of candidate passages, and a computationally expensive reader that examines the passages to identify the correct answer. Previous works have shown that as the number of retrieved passages increases, so does the performance of the reader. However, they assume all retrieved passages are of equal importance and allocate the same amount of computation to them, leading to a substantial increase in computational cost. To reduce this cost, we propose the use of adaptive computation to control the computational budget allocated for the passages to be read. We first introduce a technique operating on individual passages in isolation which relies on anytime prediction and a per-layer estimation of an early exit probability. We then introduce SKYLINEBUILDER, an approach for dynamically deciding on which passage to allocate computation at each step, based on a resource allocation policy trained via reinforcement learning. Our results on SQuAD-Open show that adaptive computation with global prioritisation improves over several strong static and adaptive methods, leading to a 4.3x reduction in computation while retaining 95{\%} performance of the full model.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wu-etal-2020-dont-read">
<titleInfo>
<title>Don’t Read Too Much Into It: Adaptive Computation for Open-Domain Question Answering</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuxiang</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pasquale</namePart>
<namePart type="family">Minervini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pontus</namePart>
<namePart type="family">Stenetorp</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Riedel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2020-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of SustaiNLP: Workshop on Simple and Efficient Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nafise</namePart>
<namePart type="given">Sadat</namePart>
<namePart type="family">Moosavi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Angela</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vered</namePart>
<namePart type="family">Shwartz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Goran</namePart>
<namePart type="family">Glavaš</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shafiq</namePart>
<namePart type="family">Joty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alex</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thomas</namePart>
<namePart type="family">Wolf</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Most approaches to Open-Domain Question Answering consist of a light-weight retriever that selects a set of candidate passages, and a computationally expensive reader that examines the passages to identify the correct answer. Previous works have shown that as the number of retrieved passages increases, so does the performance of the reader. However, they assume all retrieved passages are of equal importance and allocate the same amount of computation to them, leading to a substantial increase in computational cost. To reduce this cost, we propose the use of adaptive computation to control the computational budget allocated for the passages to be read. We first introduce a technique operating on individual passages in isolation which relies on anytime prediction and a per-layer estimation of an early exit probability. We then introduce SKYLINEBUILDER, an approach for dynamically deciding on which passage to allocate computation at each step, based on a resource allocation policy trained via reinforcement learning. Our results on SQuAD-Open show that adaptive computation with global prioritisation improves over several strong static and adaptive methods, leading to a 4.3x reduction in computation while retaining 95% performance of the full model.</abstract>
<identifier type="citekey">wu-etal-2020-dont-read</identifier>
<identifier type="doi">10.18653/v1/2020.sustainlp-1.9</identifier>
<location>
<url>https://aclanthology.org/2020.sustainlp-1.9</url>
</location>
<part>
<date>2020-11</date>
<extent unit="page">
<start>63</start>
<end>72</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Don’t Read Too Much Into It: Adaptive Computation for Open-Domain Question Answering
%A Wu, Yuxiang
%A Minervini, Pasquale
%A Stenetorp, Pontus
%A Riedel, Sebastian
%Y Moosavi, Nafise Sadat
%Y Fan, Angela
%Y Shwartz, Vered
%Y Glavaš, Goran
%Y Joty, Shafiq
%Y Wang, Alex
%Y Wolf, Thomas
%S Proceedings of SustaiNLP: Workshop on Simple and Efficient Natural Language Processing
%D 2020
%8 November
%I Association for Computational Linguistics
%C Online
%F wu-etal-2020-dont-read
%X Most approaches to Open-Domain Question Answering consist of a light-weight retriever that selects a set of candidate passages, and a computationally expensive reader that examines the passages to identify the correct answer. Previous works have shown that as the number of retrieved passages increases, so does the performance of the reader. However, they assume all retrieved passages are of equal importance and allocate the same amount of computation to them, leading to a substantial increase in computational cost. To reduce this cost, we propose the use of adaptive computation to control the computational budget allocated for the passages to be read. We first introduce a technique operating on individual passages in isolation which relies on anytime prediction and a per-layer estimation of an early exit probability. We then introduce SKYLINEBUILDER, an approach for dynamically deciding on which passage to allocate computation at each step, based on a resource allocation policy trained via reinforcement learning. Our results on SQuAD-Open show that adaptive computation with global prioritisation improves over several strong static and adaptive methods, leading to a 4.3x reduction in computation while retaining 95% performance of the full model.
%R 10.18653/v1/2020.sustainlp-1.9
%U https://aclanthology.org/2020.sustainlp-1.9
%U https://doi.org/10.18653/v1/2020.sustainlp-1.9
%P 63-72
Markdown (Informal)
[Don’t Read Too Much Into It: Adaptive Computation for Open-Domain Question Answering](https://aclanthology.org/2020.sustainlp-1.9) (Wu et al., sustainlp 2020)
ACL