NepLab-IUSS/fertility-corpus

Dataset Card for FerTime

FerTime is a diachronic corpus spanning fertility-related documents from 1770 to the present, drawn from both non-clinical sources (historical and contemporary newspapers) and clinical sources (biomedical literature and medical reports).

Dataset Details

Dataset Description

FerTime was constructed by aggregating and filtering nine openly available corpora spanning newspaper archives and biomedical/clinical literature published between 1770 and the present day. The goal of the dataset is to support diachronic and cross-domain analysis of how fertility, infertility, and related reproductive-health concepts have been discussed in both lay/journalistic and clinical/scientific discourse over more than two centuries.

• Curated by: [Agnese Lombardi — NepLab, IUSS Pavia]
• Language(s) (NLP): English (en)
• **License: cc-by-nc-4.0

Data Collection and Processing

Documents were aggregated from nine source datasets (listed above) and filtered using a predefined, regex-expanded keyword list covering clinical terminology, lay/journalistic vocabulary, and archaic or literary expressions related to fertility, drawing on prior terminological work in reproductive medicine. The keyword list includes: fertility, infertility, sterility, barrenness, fecundity, conception, pregnancy, childlessness, miscarriage, womb, uterus, ovary, ovulation, semen, sperm, birth rate, procreation, reproduction, natality, childbearing, progeny, offspring, heir, fallow, unfruitful, bear children, blessed with children, denied children, curse of barrenness. Regular expressions captured inflectional and derivational variants of these terms. The overall retention rate after filtering was approximately 2.79% of the original aggregated corpus.

Table: Corpus size before and after keyword-based filtering

Source	Pre-filtering	Post-filtering	% Retained
American Stories	272,559,714	3,563,938	1.31%
US-PD-Newspapers	21,226,700	4,259,736	20.07%
PubMed	13,803,820	529,185	3.83%
HMD Newspapers	1,494,729	149,855	10.03%
Elsevier Open Access	432,266	114,262	26.43%
CNN/DailyMail	287,107	16,650	5.80%
BMJ (Internet Archive)	7,380	6,424	87.05%
MOH Reports (Wellcome)	694	583	84.01%
Total	309,812,410	8,640,633	2.79%

Retention rates vary substantially by source, reflecting differences in genre and topical focus: clinical and specialized archives with an inherent focus on health/reproduction (BMJ, MOH Reports, Elsevier Open Access) show much higher retention than general-purpose newspaper and news corpora (American Stories, CNN/DailyMail), where fertility-related content is comparatively rare relative to overall volume.

Dataset Sources

Source	Link
American Stories	https://huggingface.co/datasets/dell-research-harvard/AmericanStories
US-PD-Newspapers	https://huggingface.co/datasets/PleIAs/US-PD-Newspapers
PubMed (abdelmo/pubmed-dataset)	https://huggingface.co/datasets/abdelmo/pubmed-dataset
HMD Newspapers	https://huggingface.co/datasets/biglam/hmd_newspapers
Elsevier Open Access	https://huggingface.co/datasets/SBU-BBL/Elsevier-OpenAccess
CNN/DailyMail	https://huggingface.co/datasets/abisee/cnn_dailymail
BMJ (Internet Archive)	https://archive.org/advancedsearch.php
MOH Reports ("London's Pulse")	https://wellcomecollection.org/collections

• Paper: [Soon Available]

Licensing Information

FerTime aggregates text from nine source datasets/archives, each carrying its own license. Because the aggregated corpus is a derivative work, FerTime's overall license is constrained by the most restrictive license among its sources — in this case, the Medical Officer of Health (MOH) reports, which are released under CC BY-NC 4.0. Accordingly, FerTime is released under CC BY-NC 4.0 (non-commercial use, with attribution).

Source	License	Notes
American Stories	CC-BY 4.0	Derived from public-domain Chronicling America scans (Library of Congress)
US-PD-Newspapers	CC0 1.0	Public domain (Library of Congress, Chronicling America)
PubMed (abdelmo/pubmed-dataset)	Unspecified	No license/dataset card provided by the source repo. Users should additionally consult NLM's PubMed usage policy, as underlying abstracts may be copyrighted by original publishers
HMD Newspapers	CC0 1.0	Public-domain 19th-century British Library newspapers
Elsevier Open Access	Unspecified / mixed	No license metadata provided on the aggregation's HF repo. Underlying Elsevier OA articles are published under a mix of terms set by the publisher/funder (e.g., CC BY, CC BY-NC, or CC BY-NC-ND), not a single uniform license
CNN/DailyMail	Apache-2.0	Note: v3.0.0 (used here) is not anonymized — individual names may appear in the text
BMJ (Internet Archive)	Uncertain	Digitized scans of historical BMJ issues. Pre-1930 issues are likely in the public domain in the US, but Internet Archive's periodicals collection is not uniformly openly licensed; status should be verified volume-by-volume
MOH Reports ("London's Pulse")	CC BY-NC 4.0	Requires attribution; non-commercial use only. This is the most restrictive license in the aggregation and determines FerTime's overall license

Known limitations regarding licensing

• The PubMed and Elsevier Open Access source datasets do not provide explicit, verifiable licensing information at the aggregation level. Users intending to redistribute or use FerTime commercially should independently verify the licensing status of documents drawn from these two sources before doing so.
• The BMJ subset was retrieved from the Internet Archive rather than from a source with an explicit open license; its public-domain status should be confirmed on a per-volume basis, particularly for issues published after 1930.
• Because FerTime is a filtered aggregation (not a full redistribution) of each source, users requiring the original full corpora, or wishing to use a subset under a more permissive license (e.g., only the CC0/CC-BY sources), can filter FerTime using the source metadata field.

Citation

BibTeX:

[Soon Available]

APA:

[Soon Available]