pos-tags, "muss"

[mxi-hug]

small observation on pos-tags:

count("GERMAPARL", query = '"muss"', cqp = T, breakdown = T, p_attribute = "pos")

results in ~30k pos = "NN", which is about 20% of all hits.

As I'm not familiar with the pos-tagger, i've no idea, whether it is possible or feasible to optimize the results...

[ablaette]

This issue is old ... but still relevant. I just inspected the the NN-tagged "muss" matches using this snippet:

k <- kwic("GERMAPARL", query = '[word = "muss" & pos = "NN"]', cqp = T)

I would hope that our pos tagger performed better than that. We have started to use StanfordNLP - and need to inspect these POS tags.