how to add header when using a parser I trained rather than provided by conllu

[toufiglu]

Hi,
I am trying to convert the spacy doc into conllu. I saw on https://pypi.org/project/spacy-conll/ that if I use an existing parser like en_core_web_sm, I can set up the header by using init_parser (include_headers=True), like the example on https://spacy.io/universe/project/spacy-conll :

nlp = init_parser("en",
                  "stanza",
                  parser_opts={"use_gpu": True, "verbose": False},
                  include_headers=True)

But how to do the same if I trained a parser myself and want to include a header? The document does not seem to have explicitly said how to do so, like in:

import numpy
from numpy import savetxt
doc = nlp ("كنت كتعصب فاش مكتبغيش تصيب ليك")
print (doc._.conll_pd)

id     form      lemma upostag     xpostag  \
0   1      كنت        كنت    NOUN  N------S2R   
1   1    كتعصب  كَتَعصبَة    NOUN  N------S2D   
2   1      فاش       فُاش    NOUN  N------S2D   
3   2  مكتبغيش    مكتبغيش    NOUN  N------S2D   
4   3     تصيب      تصيِب     ADP  P---------   
5   4      ليك       لُيك    NOUN  N------S2I   

                                feats  head deprel deps           misc  
0  Case=Gen|Definite=Cons|Number=Sing     0   ROOT    _              _  
1   Case=Gen|Definite=Def|Number=Sing     0   ROOT    _              _  
2   Case=Gen|Definite=Def|Number=Sing     0   ROOT    _              _  
3   Case=Gen|Definite=Def|Number=Sing     1   nmod    _              _  
4                        AdpType=Prep     2   nmod    _              _  
5   Case=Gen|Definite=Ind|Number=Sing     3   amod    _  SpaceAfter=No 

Thank you so much!

[BramVanroy]

Hi @toufiglu thanks for the question. If you load your own spaCy model, you'll need to add the spacy_conll component separately, like so:

nlp = spacy.load(...)  # load your own trained parser

# add spacy_conll
config = {"include_headers": True}
nlp.add_pipe("conll_formatter", config=config, last=True)

# parse
doc = nlp ("كنت كتعصب فاش مكتبغيش تصيب ليك")
print (doc._.conll_pd)

This is described in Usage.

[toufiglu]

Hi, Bram Thank you so much for answering!!! oh, it is this line. Sorry, I am new to programming. will do better!

[BramVanroy]

No problem! Make sure to mark the answer above as "Mark as answer" if your problem is solved.