A Multilingual Social Media Linguistic Corpus
Luis Rei, Dunja Mladenić, Simon Krek · 2016
This paper focuses on multilingual social media and introduces the xLiMe Twitter Corpus that contains messages in German, Italian and Spanish manually annotated with Part-of-Speech, Named Entities, and Message-level sentiment polarity. In total, the corpus contains almost 20K annotated messages and 350K tokens. The corpus is distributed in language specific files in the tab-separated values format. It also includes scripts that enable to convert sequence tagging tasks to a format similar to the CONLL format. Tokenization and pre-tagging scripts are distributed together with the data.