Photo credits
From the Rosetta Stone to punched-card stores and typewriter keys, every photograph on Parallel Text is public domain or CC0, sourced from Wikimedia Commons. Several were produced by the US federal government (National Archives, US Army, National Park Service) or are held by the Library of Congress. The only processing was a 3:2 crop, a resize and conversion to WebP. No image, logo or graphic from the MOUSSE project website that used this domain appears here.
These pictures are visual metaphors for language technology. The interpreters, programmers, visitors and places in them have no connection with Parallel Text and endorse none of its articles.
Photos by page
- Home page (top): Museum visitors leaning on a rail to look at the Rosetta Stone, a dark inscribed slab displayed at an angle. Photo: Rick Dikeman, Rosetta-stone-display-in-1985.jpg (Wikimedia Commons), CC0.
- Bias in NLP: Where It Comes From, How It Is Measured and What Helps: Two women programmers set up the wiring panels of the room-sized ENIAC computer in a black-and-white photograph. Photo: Unidentified U.S. Army photographer, Reprogramming ENIAC.png (Wikimedia Commons), Public domain.
- Benchmark Data Contamination: When Test Sets Leak into Training Data: Students sitting alone at separate white desks in a university classroom, writing answers on exam papers. Photo: KF, Test (student assessment).jpeg (Wikimedia Commons), Public domain.
- Dataset Licences for NLP: Creative Commons, Open Data Commons and the EU Rules: Rows upon rows of stacked boxes of punched record cards filling a large warehouse floor. Photo: US National Archives, IBM card storage.NARA.jpg (Wikimedia Commons), Public domain.
- Datasheets, Data Statements and Model Cards: Documenting NLP Data and Models: An old bound laboratory logbook lying open on a desk, its handwritten pages yellowed and stained. Photo: Daderot, Lab book - Lawrence Berkeley National Laboratory - DSC08822.JPG (Wikimedia Commons), Public domain.
- Language Technology in Europe: 24 Official Languages and the Digital Divide: A street at the entrance of a small town with half-timbered houses, a banner over the road and a stone gate bearing a bilingual sign. Photo: Ulamm, S-Jean-Pied-de-P Donibane.png (Wikimedia Commons), Public domain.
- How NLP Models Are Evaluated: Benchmarks, Metrics and Their Blind Spots: A yellow tape measure marked in feet and centimetres lying on a wooden floor next to two copper coins. Photo: Stilfehler, Measuring Tape Inch+CM.jpg (Wikimedia Commons), Public domain.
- Memorisation in Language Models: How Training Data Leaks and How It Is Measured: A yellow tape measure marked in feet and centimetres lying on a wooden floor next to two copper coins. Photo: Stilfehler, Measuring Tape Inch+CM.jpg (Wikimedia Commons), Public domain.
- GDPR and Text Data: Personal Data in NLP Corpora and Models: A rusty metal hasp and padlock fastening a weathered red wooden door. Photo: Estormiz, Padlock Valimokatu Oulu 20250307 01.jpg (Wikimedia Commons), CC0.
- How Tokenization Works: Words, Subwords and Bytes in NLP: Close-up black-and-white photograph of round typewriter keys arranged in staggered rows. Photo: Ralph Steiner (Clarence H. White School of Photography), Typewriter keys LCCN2004666344.jpg (Wikimedia Commons), Public domain.
- Language Identification: How NLP Systems Tell Which Language a Text Is In: Blue street-name plates on the corner of a weathered wooden house in Saransk, one giving the same street name in Russian, Erzya and Moksha. Photo: SarRus, Trilingual street sign in Saransk.jpg (Wikimedia Commons), Public domain.
- Low-Resource Languages in NLP: What "Low-Resource" Means and What Works: A wooden printer's type case full of metal type pieces for the Cherokee syllabary, scattered in its compartments. Photo: National Trails Office (US National Park Service), Historic Cherokee printing press letters at New Echota in Gordon County, Georgia (9fb7abb0-05d3-415b-a671-839f2fd5505a).JPG (Wikimedia Commons), Public domain.
- Machine Translation Evaluation: BLEU, chrF, COMET and Human Judgement: Two men wearing headphones sit at a desk with microphones and an amplifier, working as interpreters in a black-and-white photograph. Photo: Unknown author, German War Crimes Trials. Nuernberg & Dachau - NARA - 292602 (cropped).jpg (Wikimedia Commons), Public domain.
- Machine Translation for Tenants Moving Abroad: What Works and What to Check: A row of three-storey brick apartment houses with bay windows and small rental signs in the windows, in an old black-and-white photograph. Photo: Unknown author, Residential buildings with "For Rent" signs on an unidentified street.jpg (Wikimedia Commons), Public domain.
- Multilingual Lexical Knowledge Bases: WordNets, BabelNet and Wikidata Explained: The domed main reading room of the Library of Congress, with concentric rings of desks around a central circulation desk. Photo: Carol M. Highsmith, LOC Main Reading Room Highsmith.jpg (Wikimedia Commons), Public domain.
- Named Entity Recognition: How NLP Systems Find People, Places and Organisations: A metal signpost on a town square, its white arms pointing to Moscow, Washington, Lucca, Győr and other places with their distances in kilometres, under a blue sky with flags. Photo: Hearty chortle, Colmar European Cities Signpost and Flags.jpg (Wikimedia Commons), CC0.
- How Many Languages Does Machine Translation Cover? 464 Across Nine Systems, About 6% of Living Languages: An etching of the Tower of Babel as a vast circular tower of stacked arcades rising above a city and a harbour. Etching by Anton Joseph von Prenner after Pieter Bruegel the Elder, 1728, Rijksmuseum, Toren van Babel, RP-P-1933-290.jpg (Wikimedia Commons), CC0.
- Parallel Corpora: How Translated Text Trains Machine Translation: A printed notice taped to a glass door announcing holiday closing dates in English, with a greeting in Basque below. Photo: Terfili, Basque notice in Boise, Idaho.JPG (Wikimedia Commons), Public domain.
- Reproducibility in NLP: Why Results Fail to Repeat and How to Fix It: Two handwritten pages of a nineteenth-century laboratory notebook with a small sketch of apparatus and dated entries. Photo: Alexander Graham Bell, AGBell Notebook.jpg (Wikimedia Commons), Public domain.
- Web-Scraped Corpora: Copyright, Consent and Opt-Outs in NLP: A server rack packed with stacked machines, blue status displays and bundles of white network cables. Photo: Derrick Coetzee, Rear of rack at NERSC data center - closeup.jpg (Wikimedia Commons), CC0.
- What Makes a Language Model "Open"? Weights, Data, Code and Licences: A technician in a checked shirt crouches with a laptop in front of an open server rack in a data centre. Photo: Derrick Coetzee, Technician with laptop working on server rack at NERSC.jpg (Wikimedia Commons), CC0.
- Word Sense Disambiguation: How Machines Decide What a Word Means: A metal drawer full of index cards from a card catalogue, resting on a wooden table. Photo: Michael Holley, Copyright Card Catalog Drawer.jpg (Wikimedia Commons), Public domain.
Typefaces
Parallel Text uses Crimson Pro, designed by Jacques Le Bailly, and Space Grotesk, designed by Florian Karsten. Both fonts are released under the SIL Open Font License 1.1 and served with the site's own files.