missing TRB feature?

[mjgaughan]

Hello! I was wondering if the following line is supposed to be commented out?

biberplus/biberplus/tagger/tag_frequencies.py

Line 80 in 0436999

# tag_counts['RB'].append(calculate_total_adverbs(tagged_df))

I think that this comment may result in a bug where the produced vector of features is one feature shorter than the list specified in the Neurobiber/biberplus paper. While the feature "RB" is produced, "TRB" is not. This seems to occur even though both RB and TRB are listed as a generated features. I may very well be reading this wrong, just wanted to ask!

[lmmx]

Looks to me like the fix would be to rename the doc-level feature to TRB in DOC_TAGS and then uncomment this line that updates it as such

I suspect this was a temporary fix (fbc5da7)

Previously it was an OOP attribute on a WordTagger class here:

biberplus/bibermda/tagger/biber_tagger.py

Line 24 in 3a55e8b

self.adverb_count = 0

...but even then it was still called RB at the doc level:

biberplus/bibermda/reducer/constants.py

Line 110 in 3a55e8b

'RB'

...in fact from looking at this point in time, it looks like the attribute was never read, instead the total adverb count came from a upos column in the data directly:

biberplus/bibermda/reducer/corpus_statistics.py

Lines 103 to 104 in 3a55e8b

	def calculate_total_adverbs(tagged_df):
	return len(tagged_df[tagged_df['upos'] == 'ADV'])

... so there was some mismatch here, looks like a long-term bug