Use custom Keyword model for plotting

[AnubhavCR7]

Hello,
I am using a custom keyword extraction model that will extract top 3 keywords from a review text. Also I have the Sentiment associated with each review. Now I am using the scattertext plot to produce the association of the keywords with the Sentiment. In the parsed_col parameter, I am passing a column (named as key_text) which is formed by concatenating the 3 keywords extracted from each review, not the actual review text column. The reason being I want to display those keywords that have been extracted by the custom model, and not by Spacy itself. But it seems like that spacy is running it's own pipeline and displaying those keywords that it is extracting from the key_text column. Below is the code snippet :

def term_freq_parsed(data):

punc_list = [ele for ele in string.punctuation]
num_list = [str(i) for i in range(1001)]
rem_list = [ele for ele in num_list]

for ele in punc_list:
    rem_list.append(ele)

rem_list.extend(list(nlp.Defaults.stop_words))

data['key_text'] = data['key_text'].apply(lambda x: nlp(x))
          
corpus = scattertext.CorpusFromParsedDocuments(data, category_col = 'Sentiment', parsed_col = 'key_text').build().remove_terms(rem_list, ignore_absences = True)

return corpus

Is there any way to make spacy use the set of keywords that the custom keyword model is extracting, and produce the Corpus accordingly ? I don't want spacy to extract terms from the key_text column on it's own, rather use the terms already provided and produce the Corpus.

Thanks and Regards.