Two data papers, one corpus.
MEDIATEXT is documented in two companion data papers — one for Central-Eastern Europe, one for the New York Times — both currently in editorial review. Each describes the same sentence-level coding scheme: CAP policy topic, sentiment, and named entities, with emotion labels for the NYT release.
- Sebők, M., Babolcsay, B., Kiss, M., Baumgartner, Z., Ring, O., Lehoczki, R., & Molnár, C. (forthcoming). MediaText Central-Eastern Europe (CEE): A Multilingual, Sentence-Level Database of over 1.4 million Articles with Policy Topics, Sentiment, and Named Entities. Media and Communication (Cogitatio Press) — currently in revision after peer review (revised manuscript due 28 May 2026).
- Sebők, M., Kiss, M., Babolcsay, B., Baumgartner, Z., Ring, O., & Tokovics, D. (forthcoming). MediaText NYT: A 174-year sentence-level database of New York Times articles (1851–present) with policy topics, sentiment, named entities, and emotion labels. Two access levels — public metadata + binary text-availability indicators on the open web, with full word embeddings on a Harvard Dataverse + OSF deposit for the 1851–2024 closed period.
- Sebők, M., Proksch, S.-O., Rauh, C., Visnovitz, P., Balázs, G., & Schwalbach, J. (2023). Comparative computational text analysis of European legislatures: A review. International Political Science Review.
Reading portal
The V4 release is browseable at mediatext.poltextlab.com. The portal exposes an Overview, a per-country Sentiment distribution, NER word clouds, a per-row Data Explorer, and a Dashboard Information panel that documents column names against the paper codebook. A reading portal for the NYT release is being prepared on the same template.
Open data
Each release is distributed in two access tiers:
- Public — title-level metadata plus binary text-availability indicators (has_title, has_abstract, …) on the reading portal and on Harvard Dataverse.
- Programmatic / gated — vector embeddings on every text variable for researchers needing machine-readable full-text access. Same Dataverse + OSF deposit, gated for the closed period.
How to cite
Sebők, M., Babolcsay, B., Kiss, M., Baumgartner, Z., Ring, O., Lehoczki, R., & Molnár, C. (forthcoming). MediaText Central-Eastern Europe (CEE): A Multilingual, Sentence-Level Database of over 1.4 million Articles. Media and Communication.
Please also credit the originating publisher as the source of the underlying article text.
License
The MEDIATEXT releases are distributed under Creative Commons BY-NC 4.0. The reading portal and accompanying website are pilot infrastructure and may change without notice; the deposited datasets are versioned on Harvard Dataverse.
Working with the data?
If you have published using MEDIATEXT, or are starting work that would build on it, write to [email protected]. External research groups already in contact (University of Konstanz, others) receive embedding-tier access on request once their use case is documented and the underlying data papers are in print.