An unambiguous POS tagging set

This data set contains 1,123 short, POS-tagged sentences (extracted from an earlier data set; see https://zenodo.org/records/7694423), using the Universal tag set. The sentences can easily be POS tagged by a human tagger. However, standard POS taggers struggle with these sentences. The data file contains a header row that describes each column. The first column indicates the type of sentence (either a transcript of spoken text (0) or a sentence originating from written text (1)), the second column contains the actual sentence, with ground truth POS tags. The third column indicates the index of the mistagged token, and the remaining five columns show the tags assigned (of which at least one is a mistagging) of five different taggers.