Dataset on genre and news topic persistence between versions 1 and 2 of CLASSLA-web

Name: Dataset on genre and news topic persistence between versions 1 and 2 of CLASSLA-web
License: https://creativecommons.org/licenses/by-sa/4.0/

Ljubešić, Nikola; Dinić, Vuk; Rupnik, Peter

Dataset on genre and news topic persistence between versions 1 and 2 of CLASSLA-web

CLARIN.SI data & tools

Avtorji: Ljubešić, Nikola ; Dinić, Vuk and Rupnik, Peter

Identifikator vnosa: http://hdl.handle.net/11356/1972

URL projekta: https://clarinsi.github.io/classla-web/

Datum objave: 2026-10-13

Vrsta: corpus, text

Velikost: 38057171 entries

Jezik(i): Bosnian , Bulgarian , Croatian , Macedonian , Montenegrin , Serbian , Slovenian

Opis: The dataset contains metadata on documents from the CLASSLA-web v2 web crawl (http://hdl.handle.net/11356/2079), describing each document by its (1) document identifier, (2) URL, (3) web domain, (4) predicted genre, (5) predicted topic, (6) the length in terms of number of characters and, most importantly, (7) whether the document is considered to have a near-duplicate in the CLASSLA-web v1 corpus collection. This dataset originated from the paper “What Persists across Web Crawls? Genre and News Topic Variation in South Slavic Web Corpora”, accepted for publication at the 13th Web-as-Corpus Workshop (WaC-13), co-located with EMNLP 2026. We make the dataset openly available to facilitate follow-up research on this topic.