DLT-Corpus: A Large-Scale Text Collection for the Distributed Ledger Technology Domain

Walter Hernandez Cruz; Peter Devine; Nikhil Vadgama; Paolo Tasca; Jiahua Xu

cs.CLFeb 25, 2026

DLT-Corpus: A Large-Scale Text Collection for the Distributed Ledger Technology Domain

Authors: Walter Hernandez Cruz, Peter Devine, Nikhil Vadgama, Paolo Tasca, Jiahua Xu

Organizations: Centre for Blockchain Technologies, University College London · Exponential Science Foundation · School of Informatics, University of Edinburgh

Abstract

We introduce DLT-Corpus, the largest domain-specific text collection for Distributed Ledger Technology (DLT) research to date: 2.98 billion tokens from 22.12 million documents spanning scientific literature (37,440 publications), United States Patent and Trademark Office (USPTO) patents (49,023 filings), and social media (22 million posts). Existing Natural Language Processing (NLP) resources for DLT focus narrowly on cryptocurrency price prediction and smart contracts, leaving domain-specific language underexplored despite the sector's ~$3 trillion market capitalization and rapid technological evolution. We demonstrate DLT-Corpus' utility by analyzing patterns of technology emergence and market-innovation correlations. Findings reveal that technologies first appear in our scientific literature subset before reaching patents and social media, following traditional technology transfer patterns. While social media sentiment remains overwhelmingly bullish even during crypto winters, scientific and patent activity grows less tied to short-term sentiment, tracking overall market expansion in a virtuous cycle in which research precedes and enables economic growth that, in turn, funds further innovation. We release the DLT-Corpus and companion artifacts: LedgerBERT (+23% over BERT-base on DLT-specific Named Entity Recognition (NER) task), a sentiment analysis dataset of 23,301 crypto news headlines and descriptions, tools, and code.

Figures & tables

Panel of Figure 1 — Figure 1. Overview of Distributed Ledger Technology-Corpus composition

Corpus	Type	Density	Doc. Cov.	JS Div.
Social Media	Domain	86.92	96.3%	0.44
Patents	Domain	25.27	99.8%	0.45
Scientific Lit.	Domain	19.68	100.0%	0.39
DLT-Corpus avg.	Domain	43.96	98.7%	0.43
RefinedWeb	General	4.96	55.6%	0.10
C4	General	5.14	51.9%	0.10

Table 1. Comparison of Distributed Ledger Technology-Corpus with RefinedWeb ( Penedo et al., 2023 ) and C4 ( Raffel et al., 2019 )

General Baselines
Model	Parameters	NER F1
BERT-base ( Devlin et al., 2019 )	110M	0.243
MosaicBERT-base ( Portes et al., 2023 )	109M	0.076
ModernBERT-base ( Warner et al., 2025 )	149M	0.209
Domain-Adapted Baselines
SciBERT ( Beltagy et al., 2019 )	110M	0.289

Table 2. Performance on in-domain Distributed Ledger Technology entity recognition from scientific literature. Named Entity Recognition scores represent F1-average across 5-fold cross-validation with strict entity-level matching (exact boundary and type agreement required).

Model	Market	Content	Engagement
	Direction	Characteristics	Quality
BERT-base	0.577	0.514	0.539
DistilBERT-base	0.570	0.508	0.544
SciBERT	0.591	0.514	0.532
LedgerBERT	0.590	0.502	0.537
vs. SciBERT	-0.2%	-2.3%	+0.9%

Table 3. Performance on out-of-domain sentiment analysis from cryptocurrency news articles titles and brief descriptions.

Field	Description
paperId	Semantic Scholar identifier
title	Publication title
abstract	Publication abstract
text	Full text in Markdown format
url	URL to the source document
year	Publication year

Table 4. Fields in Scientific Literature dataset

Field	Description
Patent Number	USPTO patent number (e.g., US10123456B2)
Document ID	Unique document identifier
Title	Patent title
text	Full patent text
Date Published	Publication date
Filing Date	Application filing date

Table 5. Fields in Patents dataset

Field	Description
tweet	Tweet text
timestamp	Post creation time
year	Post year
language	Detected language
sentiment_class	Sentiment category
sentiment_label	Sentiment label

Table 6. Fields in Social Media dataset

Document Type	Period	Spearman’s $\rho$	$p$ -value
Scientific literature	2013-2024	0.76	$<$ 0.004
Patents	2013-2024	0.96	$<$ 0.001
Social media	2013-2023	0.98	$<$ 0.001

Table 7. Correlation between cryptocurrencies’ market capitalization and document volumes ( footnote 36 ).

Document volumes lead market
Lag (years)	Scientific	Social Media	Patents
$-5$	0.86**	0.77	0.86**
$-4$	0.90**	0.82*	0.90**
$-3$	0.93***	0.95***	0.93***
$-2$	0.95***	0.88**	0.95***
$-1$	0.94***	0.90***	0.95***

Table 8. Lagged correlations (Spearman’s

\rho

) between document volumes and market capitalization.

Panel of Figure 4 — Figure 4. Proportion of mentions per year for selected cryptocurrencies in the Distributed Ledger Technology-Corpus. The y-axis represents the relative frequency of mentions normalized by the total volume of documents in each source per year.

Figure 5. Market sentiment in social media and yearly growth of global cryptocurrency market capitalization.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Field	Description
title	Article headline
description	Article summary
text	Combined title and description
timestamp	Publication time
url	Original article link
source_url	CryptoPanic news link

Appendix

Table 9. Fields in Sentiment Analysis dataset

Are Whitepaper Claims Reflected in Market Structure? A Contamination-Aware Pipeline and a Power-Limited Null
Jan 28, 2026Murad FarzullaData Contamination Representation Alignment
Dissensus, London, UK · King’s College London, London, UK
SoK: A Systematic Bidirectional Literature Review of AI & DLT Convergence
May 11, 2026Ali Irzam Kathia, Yimika Erinle, Abylay Satybaldy +3
Exponential Science Foundation, Lugano, Switzerland
GPT-NL Public Corpus: A Permissively Licensed, Dutch-First Dataset for LLM Pre-training
Apr 1, 2026Jesse van Oort, Frank Brinkkemper, Erik de Graaf +2Language Model Pretraining Low-Resource Language Processing
TNO, The Netherlands · GPT-NL, The Netherlands · Instituut voor de Nederlandse Taal, The Netherlands +1

Jan 28, 2026J/K move · Enter open · S save

Are Whitepaper Claims Reflected in Market Structure? A Contamination-Aware Pipeline and a Power-Limited Null

Murad Farzulla

Dissensus, London, UK · King’s College London, London, UK

Data Contamination Representation Alignment Finance

Open paper AskPDF arXiv

Do the functional narratives in cryptocurrency whitepapers correspond to how their tokens behave in markets? We develop a content-verified, contamination-aware pipeline for measuring structural correspondence between project narratives and market structure, and report two results. The first is a cautionary one. An apparent entity-level signal in an earlier version of our corpus -- specialised tokens appearing to align more strongly than broad infrastructure tokens -- was entirely an artifact of corpus contamination: roughly a quarter of the documents were failed-download stubs or wrong-document whitepapers (for example, a "Cosmos" entry that was in fact Binance Smart Chain text), and the apparent ordering does not survive content verification: on the clean corpus no token registers as helping alignment. We therefore report it as a contamination diagnosis, not a finding. The second is an honest null. Combining zero-shot NLP classification of 43 content-verified whitepapers across 10 semantic categories with seven cross-sectional market-structure statistics computed from hourly data (17,543 timestamps, 2023-2024), and aligning the two spaces with Procrustes rotation and Tucker's congruence coefficient (

φ

), we do not detect a significant claims-market alignment in this

n = 43

sample (dimension-matched

φ= 0.303

, zero-padded

φ= 0.223

; both non-significant). A positive-control and power analysis shows the binding constraint is the low reliability of the text instrument: the minimum detectable effect is

φ\approx 0.66

, well above the observed

\approx 0.22

. This is absence of evidence for alignment, not evidence of its absence -- we can reject strong alignment (

φ\geq 0.70

) but cannot distinguish weak alignment (

φ\approx 0.3

) from none.

Abstract

Explore similar work

Are Whitepaper Claims Reflected in Market Structure? A Contamination-Aware Pipeline and a Power-Limited Null

SoK: A Systematic Bidirectional Literature Review of AI & DLT Convergence

GPT-NL Public Corpus: A Permissively Licensed, Dutch-First Dataset for LLM Pre-training