Basic reader of parc xml files

parc_reader.py

from bs4 import BeautifulSoup as Soup


class AnnotatedText(object):

	def __init__(self, parc_xml):

		self.soup = Soup(parc_xml, 'html.parser')

		self.words = []
		self.sentences = []
		sentence_tags = self.soup.find_all('sentence')
		for sentence_tag in sentence_tags:
			sentence = {'words':[]}
			self.sentences.append(sentence)
			word_tags = sentence_tag.find_all('word')
			for word_tag in word_tags:
				word = {
					'token': word_tag['text'],
				}
				attribution = word_tag.find('attribution')
				if attribution:
					word['attribution'] = {
						'role': attribution.find('attributionrole')['rolevalue'],
						'id': attribution['id']
					}
				else:
					word['attribution'] = None

				self.words.append(word)
				sentence['words'].append(word)

>>> from parc_reader import AnnotatedText as A
>>> annotated_text = A(open('some-file.xml').read())

Once you have an annotated text object, it has a property sentences which is a list of sentences. Each sentence is a dictionary, having a words property, which is a list of all the words in the sentence. Words are dictionaries too, and they have a token property (the original text of the word) and an attribution property. The attribution property can be None, or it could be a dictionary with a role and an id. The role identifies whether the word belongs to a cue, content, or source span.