The tagset used by the pipeline in all modules (pos-tagging, constituents and dependencies) is from the Ancora corpus.
However, for coarse part-of-speech tags (broad tags, without morphological details like number or gender), the pipeline uses different tags to Ancora. These broad tags are described in the table below.
The Ancora tags are also output, in the "morphofeat" attribute of the terms layer of the pipeline's NAF output (NLP Annotation Format)--see an example [here].