@inproceedings{doostmohammadi-etal-2023-surface,
title = "Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models",
author = "Doostmohammadi, Ehsan and
Norlund, Tobias and
Kuhlmann, Marco and
Johansson, Richard",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-short.45",
doi = "10.18653/v1/2023.acl-short.45",
pages = "521--529",
abstract = "Augmenting language models with a retrieval mechanism has been shown to significantly improve their performance while keeping the number of parameters low. Retrieval-augmented models commonly rely on a semantic retrieval mechanism based on the similarity between dense representations of the query chunk and potential neighbors. In this paper, we study the state-of-the-art Retro model and observe that its performance gain is better explained by surface-level similarities, such as token overlap. Inspired by this, we replace the semantic retrieval in Retro with a surface-level method based on BM25, obtaining a significant reduction in perplexity. As full BM25 retrieval can be computationally costly for large datasets, we also apply it in a re-ranking scenario, gaining part of the perplexity reduction with minimal computational overhead.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="doostmohammadi-etal-2023-surface">
<titleInfo>
<title>Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ehsan</namePart>
<namePart type="family">Doostmohammadi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tobias</namePart>
<namePart type="family">Norlund</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marco</namePart>
<namePart type="family">Kuhlmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Richard</namePart>
<namePart type="family">Johansson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Augmenting language models with a retrieval mechanism has been shown to significantly improve their performance while keeping the number of parameters low. Retrieval-augmented models commonly rely on a semantic retrieval mechanism based on the similarity between dense representations of the query chunk and potential neighbors. In this paper, we study the state-of-the-art Retro model and observe that its performance gain is better explained by surface-level similarities, such as token overlap. Inspired by this, we replace the semantic retrieval in Retro with a surface-level method based on BM25, obtaining a significant reduction in perplexity. As full BM25 retrieval can be computationally costly for large datasets, we also apply it in a re-ranking scenario, gaining part of the perplexity reduction with minimal computational overhead.</abstract>
<identifier type="citekey">doostmohammadi-etal-2023-surface</identifier>
<identifier type="doi">10.18653/v1/2023.acl-short.45</identifier>
<location>
<url>https://aclanthology.org/2023.acl-short.45</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>521</start>
<end>529</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models
%A Doostmohammadi, Ehsan
%A Norlund, Tobias
%A Kuhlmann, Marco
%A Johansson, Richard
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F doostmohammadi-etal-2023-surface
%X Augmenting language models with a retrieval mechanism has been shown to significantly improve their performance while keeping the number of parameters low. Retrieval-augmented models commonly rely on a semantic retrieval mechanism based on the similarity between dense representations of the query chunk and potential neighbors. In this paper, we study the state-of-the-art Retro model and observe that its performance gain is better explained by surface-level similarities, such as token overlap. Inspired by this, we replace the semantic retrieval in Retro with a surface-level method based on BM25, obtaining a significant reduction in perplexity. As full BM25 retrieval can be computationally costly for large datasets, we also apply it in a re-ranking scenario, gaining part of the perplexity reduction with minimal computational overhead.
%R 10.18653/v1/2023.acl-short.45
%U https://aclanthology.org/2023.acl-short.45
%U https://doi.org/10.18653/v1/2023.acl-short.45
%P 521-529
Markdown (Informal)
[Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models](https://aclanthology.org/2023.acl-short.45) (Doostmohammadi et al., ACL 2023)
ACL