what are the ￨_ ?

[arademaker]

dmrs-text-generation-naacl2019/data/sample/sample-src.txt

Line 1 in c6d9e31

(￨_ _open_v_1￨mood=INDICATIVE|perf=-|sf=PROP|tense=PAST ARG1-NEQ￨_ (￨_ _window_n_1￨ind=+|num=SG|pers=3 RSTR-H-of￨_ (￨_ _the_q￨_ )￨_ )￨_ )￨_

Do you guys remember what those ￨_ symbols are? Fig 2 in https://arxiv.org/pdf/1904.11564.pdf does not show them. I didn't find any explanation in the paper about those extra symbols before the predicates. @goodmami ?

[goodmami]

No, I don't. All I can find is this:

dmrs-text-generation-naacl2019/preprocessing.py

Lines 103 to 105 in c6d9e31

	# Tokens with "￨" were already featurized by combine_attributes. For anything else,
	# add empty feature _ to fill requirement that all tokens have same number of features.
	featurized_tokens = [token if u'￨' in token else token + u'￨_' for token in tokens]

Interestingly, that ￨ character is not a regular pipe character |, but a halfwidth vertical bar character for box drawing: https://codepoints.net/U+FFE8. I don't recall writing this code so I cannot provide any more info about its use other than the comments in the code above.