Multilingual News Corpus

Stats

• Language: English, German
• 402358 Articles
• 11 classes
• Rich meta information available for the articles

Download

1. Get Access to the Corpus

Since the news articles themselfes are copyrighted to the original publisher, access to this corpus can't be made public. If you want access, drop a short Email to griesshaber@hdm-stuttgart.de.

2. I already have access

Download the newest release here

Classes

The corpus is split into the following classes:

Label	Documents (EN)	Documents (DE)
Wirtschaft	2519	53769
Lokal	2309	18037
Finanzen	146	13339
Ausland	16972	25455
Kultur	1583	11714
Ignore	399	3488
Aktuell	2319	1253
Politik	8822	72414
Technologie	3280	25956
Sport	3540	24704
Lifestyle	13637	19476
Sonstiges	24180	53047
Total	79706	322652

Metainformation

The corpus was build by crawling german and english news sites. To extract the information unfluff was used. Therefore all metainformation is available as it is extracted by unfluff.

• rssTitle: String
• source: String
• title: String
• site: String
• softTitle: String
• date: Date
• copyright: String
• author: [String]
• publisher: String
• copyright: String
• favicon: String
• description: String
• lang: String
• canonicalLink: String
• tags: [String]
• isage: String
• videos: [[URL]]
• links: [[URL]]
• text: String

Citing

If you find this corpus helpful in your research, consider citing this work:

@misc{Griesshaber2017,
  author = {Grie{\ss}haber, Daniel},
  title = {Multilingual News Corpus},
  year = {2017},
  publisher = {GitLab HdM Stuttgart},
  journal = {Git repository},
  howpublished = {\url{https://gitlab.mi.hdm-stuttgart.de/griesshaber/german-news-corpus}}
}
      

Author

Daniel Grießhaber (Twitter/GitHub).

Template

Solo by chibicode