Spaces:
Sleeping
Sleeping
File size: 11,703 Bytes
57cf043 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 |
from typing import List, Tuple, Optional
import pandas as pd
class MetadataManager:
def __init__(self, df: pd.DataFrame, logger):
self.logger = logger
self.df = df
self.df.drop('Embedding', axis=1, inplace=True)
self.df = self.df.where(pd.notna(self.df), 'unknown')
@staticmethod
def __search_sub_level(df: pd.DataFrame, header_text: Optional[str] = None) -> List:
"""
Args:
df:
Returns:
"""
paragraphs = []
if header_text is None:
header_text = df.iloc[0]['Text']
for ind, (_, row) in enumerate(df.iterrows()):
text = row['Text']
if ind == 0:
text = text.replace(f'{header_text}', f'{header_text}\n')
else:
text = text.replace(f'{header_text}', '') + '\n'
paragraphs.append(text)
return paragraphs
@staticmethod
def __check_duplicates(df: pd.DataFrame, ind: int) -> pd.DataFrame:
if df.loc[ind]['Duplicate'] is not None:
return df[df['Duplicate'] == df.loc[ind]['Duplicate']]
else:
return df[df['Duplicate'].isna()]
@staticmethod
def __check_appendix_duplicates(df: pd.DataFrame, ind: int) -> pd.DataFrame:
if df.loc[ind]['DuplicateAppendix'] is not None:
return df[df['DuplicateAppendix'] == df.loc[ind]['DuplicateAppendix']]
else:
return df[df['DuplicateAppendix'].isna()]
def _paragraph_appendix_content(self, df, pattern: str, ind: int, shape: int) -> Tuple[List, int]:
"""
Функция возвращает контент параграфа. Если в параграфе были подпункты через "-" или буквы "а, б"
Args:
df: DataFrame
pattern: Паттерн поиска.
ind: Индекс строки в DataFrame.
shape: Размер DataFrame при котором будет возвращаться пустой список.
Returns:
Возвращает список подразделов.
Examples:
3.1. Параграф:
1) - Содержание 1;
2) - Содержание 2;
3) - Содержание 3;
"""
df = df[(df['PargaraphAppendix'].str.match(pattern, na=False)) | (df.index == ind)]
df = self.__check_appendix_duplicates(df, ind)
if df.shape[0] <= shape:
return [], None
start_index_paragraph = df.index[0]
paragraphs = self.__search_sub_level(df)
return paragraphs, start_index_paragraph
def _paragraph_content(self, df, pattern: str, ind: int, shape: int) -> Tuple[List, int]:
"""
Функция возвращает контент параграфа. Если в параграфе были подпункты через "-" или буквы "а, б"
Args:
df: DataFrame
pattern: Паттерн поиска.
ind: Индекс строки в DataFrame.
shape: Размер DataFrame при котором будет возвращаться пустой список.
Returns:
Возвращает список подразделов.
Examples:
3.1. Параграф:
1) - Содержание 1;
2) - Содержание 2;
3) - Содержание 3;
"""
df = df[
(df['Pargaraph'].str.match(pattern, na=False)) & # Проверка, соответствуют ли значения паттерну
(df['Duplicate'] == df.loc[ind]['Duplicate']) | # Оставить разделы только принадлежащие одному дубликату
(df.index == ind)] # Оставить значение, которое нашел векторный поиск
# df = self.__check_duplicates(df, ind)
if df.shape[0] <= shape:
return [], None
start_index_paragraph = df.index[0]
paragraphs = self.__search_sub_level(df)
return paragraphs, start_index_paragraph
def _paragraph_content2(self, df, pattern: str, ind: int, shape: int) -> Tuple[List, int]:
"""
Функция возвращает контент параграфа. Если в параграфе были подпункты через "-" или буквы "а, б"
Args:
df: DataFrame
pattern: Паттерн поиска.
ind: Индекс строки в DataFrame.
shape: Размер DataFrame при котором будет возвращаться пустой список.
Returns:
Возвращает список подразделов.
Examples:
3.1. Параграф:
1) - Содержание 1;
2) - Содержание 2;
3) - Содержание 3;
"""
df = df[df['Pargaraph'].str.match(pattern, na=False)]
if df.shape[0] <= shape:
return [], None
# df = self.__check_duplicates(df, ind)
# if df.shape[0] <= shape:
# return [], None
start_index_paragraph = df.index[0]
paragraphs = self.__search_sub_level(df)
return paragraphs, start_index_paragraph
@staticmethod
def _first_unknown_index(df):
indexes = list(df[df['PartLevel1'].isin(['unknown'])].index)
if len(indexes) > 0:
return df.loc[indexes[-1]]['Text']
else:
return None
def _search_other_info(self, ind, doc_number):
df = self.df[self.df['DocNumber'] == doc_number]
start_index_paragraph = df.loc[ind]['Index'] - 1
if df.loc[ind]['Table'] != 'unknown':
return df.loc[ind]['Text'], ind
if df.loc[ind]['PartLevel1'] != 'unknown':
if 'Table' in str(self.df.iloc[ind]['PartLevel1']):
return [], ind
if df.loc[ind]['Appendix'] != 'unknown':
df = df[df['Appendix'] == self.df.iloc[ind]['Appendix']]
if df.loc[ind]['LevelParagraphAppendix'] == 'unknown' and df.loc[ind]['PargaraphAppendix'] == 'unknown':
# pattern = r'\d+\.?$'
# df = df[(df['PargaraphAppendix'].str.match(pattern, na=False)) | (df.index == ind)]
# df = df[(df['LevelParagraphAppendix'] == 'Level0') | (df.index == ind)]
df = df.loc[ind:ind + 7]
start_index_paragraph = df.index[0]
paragraph = self.__search_sub_level(df)
elif df.loc[ind]['PargaraphAppendix'] != 'unknown':
pattern = df.loc[ind]["PargaraphAppendix"].replace(".", r"\.")
pattern = f'^{pattern}?\\d?.?$'
if df[df['PargaraphAppendix'].str.match(pattern, na=False)].shape[0] == 1:
pattern = df.loc[ind]["PargaraphAppendix"].replace(".", r"\.")
pattern = pattern.split('.')
pattern = [elem for elem in pattern if elem]
if len(pattern) == 1:
pattern = '.'.join(pattern)
pattern = f'^{pattern}.?\\d?.?$'
else:
pattern = '.'.join(pattern[:-1])
pattern = f'^{pattern}.\\d.?$'
df = df[df['PargaraphAppendix'].str.match(pattern, na=False)]
start_index_paragraph = df.index[0]
paragraph = self.__search_sub_level(df)
else:
paragraph = self.df.iloc[int(ind - 10):ind + 10]['Text'].values
start_index_paragraph = df.index[0]
return ' '.join(paragraph), start_index_paragraph
else:
if df.loc[ind]['Pargaraph'] == 'unknown':
header_text = self._first_unknown_index(df)
df = df.loc[int(ind - 2):ind + 2]
paragraph = self.__search_sub_level(df, header_text)
# Связан с документами без пунктов поэтому передается несколько параграфов сверху и снизу
else:
pattern = df.loc[ind]["Pargaraph"].replace(".", r"\.")
# Изет под пункты внутри пункта
paragraph, start_index_paragraph = self._paragraph_content(df, fr'^{pattern}?$', ind, 2)
if len(paragraph) == 0:
pattern = f'{pattern}\\d?.?\\d?\\d?.?$'
paragraph, start_index_paragraph = self._paragraph_content2(df, pattern, ind, 0)
if len(paragraph) == 0 and df.loc[ind]['LevelParagraph'] != '0':
pattern = df.loc[ind]["Pargaraph"].split('.')
pattern = [elem for elem in pattern if elem]
pattern = '.'.join(pattern[:-1])
pattern = f'^{pattern}\\.\\d\\d?.?$'
paragraph, start_index_paragraph = self._paragraph_content(df, pattern, ind, 0)
elif len(paragraph) == 0 and df.loc[ind]['LevelParagraph'] == '0':
pattern = df.loc[ind]["Pargaraph"].replace(".", r"\.")
if '.' not in pattern:
pattern = pattern + '\.'
pattern = f'^{pattern}\\d.?\\d?.?$'
paragraph, start_index_paragraph = self._paragraph_content(df, pattern, ind, 0)
return ' '.join(paragraph), start_index_paragraph
@staticmethod
def filter_answer(answer):
flip_answer = []
new_answer = {}
count = 0
for key in answer:
if answer[key]['start_index_paragraph'] not in flip_answer:
flip_answer.append(answer[key]['start_index_paragraph'])
new_answer[count] = answer[key]
count += 1
return new_answer
def _clear_doc_name(self, ind):
split_doc_name = self.df.iloc[ind]['DocName'].split('_')
return ' '.join(split_doc_name[1:]).replace('.txt', '').replace('.json', '').replace('.DOCX', '').replace(
'.DOC', '').replace('tables', '')
def search(self, indexes: List) -> dict:
"""
Метод ищет ответы на запрос
Args:
indexes: Список индексов.
Returns:
Возвращает словарь с ответами и информацией об ответах.
"""
answers = {}
for i, ind in enumerate(indexes):
answers[i] = {}
doc_number = self.df.iloc[ind]['DocNumber']
answers[i]['id'] = doc_number
answers[i][f'index_answer'] = int(ind)
answers[i][f'doc_name'] = self._clear_doc_name(ind)
answers[i][f'title'] = self.df.iloc[ind]['Title']
answers[i][f'text_answer'] = self.df.iloc[ind]['Text']
try:
other_info, start_index_paragraph = self._search_other_info(ind, doc_number)
except KeyError:
other_info, start_index_paragraph = self.df.iloc[ind]['Text'], ind
self.logger.info('Ошибка в индексе, проверьте БД!')
if len(other_info) == 0:
other_info, start_index_paragraph = self.df.iloc[ind]['Text'], ind
answers[i][f'other_info'] = [other_info]
answers[i][f'start_index_paragraph'] = int(start_index_paragraph)
return self.filter_answer(answers)
|