LiMoSINe pipeline: Multilingual UIMA-based NLP platform

Abstract

We present a robust and efficient parallelizable multilingual UIMA-based platform for automatically annotating textual inputs with different layers of linguistic description, ranging from surface level phenomena all the way down to deep discourse-level information. In particular, given an input text, the pipeline extracts: sentences and tokens; entity mentions; syntactic information; opinionated expressions; relations between entity mentions; co-reference chains and wikified entities. The system is available in two versions: a standalone distribution enables design and optimization of userspecific sub-modules, whereas a server-client distribution allows for straightforward highperformance NLP processing, reducing the engineering cost for higher-level tasks.