Dataset origin: https://zenodo.org/records/3829167

Description

Content:

List of the 743 domains, their term vocabularies in 10 languages, and the Wikipedia articles associated to each domain extracted by the best model described in:

Cristina España-Bonet, Alberto Barrón-Cedeño and Lluís Màrquez. "Tailoring and Evaluating the Wikipedia for in-Domain Comparable Corpora Extraction." Knowledge and Information Systems, Volume 65, pages 1365-1397. 2023. Springer-Verlag, London Ldt. https://doi.org/10.1007/s10115-022-01767-5

https://github.com/cristinae/WikiTailor

Files Description:

commonCats2015.enesdefrcaareuelrooc.tsv

Multilingual domains listed one per line, languages are separated by a tab in the order en, es, de, fr, ca, ar, eu, el, ro and oc. For each language we include the pair "ID categoryName" separated by a blank space.

[LAN].0.tar.bz

A folder per domain for language [LAN] containing the vocabulary and IDs of the extracted articles by the Wikitailor model 50-WT100.

extraction[LAN]0.tar.bz

A folder per domain for language [LAN] containing the text of the extracted articles. The name of the file corresponds to the IDs in [LAN].0.tar.bz.

Citation

España-Bonet, C., Barrón-Cedeño, A., & Màrquez, L. (2020). WTC1.1 (WikiTailor corpus v. 1.1) (v1.1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3829167