@inproceedings{garg-etal-2026-data,
title = "Data Scout: Targeted Web Crawling for Domain-Specific Pretraining Corpora",
author = "Garg, Chirag and
Zahid, Eelaaf and
Ahmed, Farhan and
Gala, Jay and
Butler, Eric K and
Ludwig, Heiko",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.31/",
pages = "519--531",
ISBN = "979-8-89176-460-6",
abstract = "The dominant approach to building domain-specific pretraining corpora is to filter large web archives such as CommonCrawl. This works well for popular domains but breaks down for specialized ones, where relevant content is sparse and often beyond the reach of popularity-driven crawlers. We present **Data Scout**, which inverts this: instead of filtering an archive, it directs a targeted crawl. An LLM expands a root topic into a taxonomy and thousands of search queries; the returned URLs (*seeds*) are grouped by subdomain and screened with a user-supplied classifier (the *probe*), admitting each subdomain on the basis of a small sample. This works because relevance has a sharp boundary at the subdomain level: in mathematics, a page is 21x more likely to be relevant than one on a sibling subdomain. With the FineMath classifier as the probe, **21.9{\%}** of crawled pages are high-quality math content, **70x** the 0.31{\%} rate from filtering a comparable web sample, so the crawl wastes far less effort. But the payoff is not just efficiency: 63.2{\%} of these pages are missing from CommonCrawl altogether, yet just as useful for training. Continued pretraining of Llama-3.2-3B on 1.9B Data Scout tokens matches FineMath corpus on GSM8k. Because the probe is the only domain-specific component, Data Scout can in principle apply to any domain with such a classifier."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="garg-etal-2026-data">
<titleInfo>
<title>Data Scout: Targeted Web Crawling for Domain-Specific Pretraining Corpora</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chirag</namePart>
<namePart type="family">Garg</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eelaaf</namePart>
<namePart type="family">Zahid</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Farhan</namePart>
<namePart type="family">Ahmed</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jay</namePart>
<namePart type="family">Gala</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eric</namePart>
<namePart type="given">K</namePart>
<namePart type="family">Butler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heiko</namePart>
<namePart type="family">Ludwig</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>The dominant approach to building domain-specific pretraining corpora is to filter large web archives such as CommonCrawl. This works well for popular domains but breaks down for specialized ones, where relevant content is sparse and often beyond the reach of popularity-driven crawlers. We present **Data Scout**, which inverts this: instead of filtering an archive, it directs a targeted crawl. An LLM expands a root topic into a taxonomy and thousands of search queries; the returned URLs (*seeds*) are grouped by subdomain and screened with a user-supplied classifier (the *probe*), admitting each subdomain on the basis of a small sample. This works because relevance has a sharp boundary at the subdomain level: in mathematics, a page is 21x more likely to be relevant than one on a sibling subdomain. With the FineMath classifier as the probe, **21.9%** of crawled pages are high-quality math content, **70x** the 0.31% rate from filtering a comparable web sample, so the crawl wastes far less effort. But the payoff is not just efficiency: 63.2% of these pages are missing from CommonCrawl altogether, yet just as useful for training. Continued pretraining of Llama-3.2-3B on 1.9B Data Scout tokens matches FineMath corpus on GSM8k. Because the probe is the only domain-specific component, Data Scout can in principle apply to any domain with such a classifier.</abstract>
<identifier type="citekey">garg-etal-2026-data</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.31/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>519</start>
<end>531</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Data Scout: Targeted Web Crawling for Domain-Specific Pretraining Corpora
%A Garg, Chirag
%A Zahid, Eelaaf
%A Ahmed, Farhan
%A Gala, Jay
%A Butler, Eric K.
%A Ludwig, Heiko
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F garg-etal-2026-data
%X The dominant approach to building domain-specific pretraining corpora is to filter large web archives such as CommonCrawl. This works well for popular domains but breaks down for specialized ones, where relevant content is sparse and often beyond the reach of popularity-driven crawlers. We present **Data Scout**, which inverts this: instead of filtering an archive, it directs a targeted crawl. An LLM expands a root topic into a taxonomy and thousands of search queries; the returned URLs (*seeds*) are grouped by subdomain and screened with a user-supplied classifier (the *probe*), admitting each subdomain on the basis of a small sample. This works because relevance has a sharp boundary at the subdomain level: in mathematics, a page is 21x more likely to be relevant than one on a sibling subdomain. With the FineMath classifier as the probe, **21.9%** of crawled pages are high-quality math content, **70x** the 0.31% rate from filtering a comparable web sample, so the crawl wastes far less effort. But the payoff is not just efficiency: 63.2% of these pages are missing from CommonCrawl altogether, yet just as useful for training. Continued pretraining of Llama-3.2-3B on 1.9B Data Scout tokens matches FineMath corpus on GSM8k. Because the probe is the only domain-specific component, Data Scout can in principle apply to any domain with such a classifier.
%U https://aclanthology.org/2026.emnlp-industry.31/
%P 519-531
Markdown (Informal)
[Data Scout: Targeted Web Crawling for Domain-Specific Pretraining Corpora](https://aclanthology.org/2026.emnlp-industry.31/) (Garg et al., EMNLP 2026)
ACL