# from ..formulated_constants import include_toggled_phrases, include_toggled_pairs
import formulated_constants
import operator
import pycrfsuite
import numpy as np
from sklearn.metrics import classification_report
import re
import os
# http://www.albertauyeung.com/post/python-sequence-labelling-with-crf/
TRAIN_FOLDER = '../annotated'
cue_phrases = formulated_constants.categorical_phrases
cue_pairs = formulated_constants.categorical_pairs
# check to include variations in case
formulated_constants.include_toggled_phrases(cue_phrases)
formulated_constants.include_toggled_pairs(cue_pairs)
category_abbr = {
"Argument":
"A",
"Fact":
"F",
"Issue":
"I",
"Ruling by lower court":
"LR",
"Ruling by the present court":
"R",
"Statute":
"SS",
# selection is for summary, based on:
# 1. position of para in doc
# 2. position of para. in thematic segment
# 3. position of sentence in paragraph
# 4. tf-idf (position of word in document and corpus)
####
# Carry out thematic segmentation first
# filtering to eliminate unimportant quotations and noises
# selection of the candidate units
# production of table style summary <= Not relevant, only restrict limit, proper grammer
cue_phrases = formulated_constants.categorical_phrases
cue_pairs = formulated_constants.categorical_pairs
# check saravanan's constants
include_toggled_phrases(cue_phrases)
include_toggled_pairs(cue_pairs)
def LetSum(file):
'''
receives a html file, produce categories for each of them and returns a summary
'''
label = OrderedDict()
scores = OrderedDict()
lines = {}
t, _ = crf_test.parse_html(file)
# break sentences to phrases
# for line in t:
# print(line)