File size: 11,261 Bytes
d9aeb25 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 |
import os
import re
import subprocess
from typing import Set, Tuple
# Third-party imports
import spacy
import jieba
import gzip
import json
from pathlib import Path
import logging
from datatrove.executor.local import LocalPipelineExecutor
from datatrove.pipeline.filters import (
RegexFilter,
LanguageFilter,
GopherQualityFilter,
C4QualityFilter,
)
from datatrove.pipeline.dedup import (
MinhashDedupCluster,
MinhashDedupFilter,
MinhashDedupSignature,
MinhashConfig,
MinhashDedupBuckets,
)
from datatrove.pipeline.writers.jsonl import JsonlWriter
from datatrove.pipeline.writers.parquet import ParquetWriter
from datatrove.pipeline.readers import ParquetReader
from datatrove.pipeline.tokens import TokensCounter
from datatrove.pipeline.formatters import PIIFormatter
# Constants
MAIN_OUTPUT_PATH = "./output"
FILTERING_OUTPUT_PATH = f"{MAIN_OUTPUT_PATH}/base_processing"
DUMP = "CC-MAIN-2024-26"
S3_MINHASH_BASE_PATH = f"{MAIN_OUTPUT_PATH}/minhash"
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
# Configuration
minhash_config = MinhashConfig(
num_buckets=14,
hashes_per_bucket=8,
n_grams=2,
)
# Chinese stop words
chinese_stop_words = [
"的", "了", "和", "是", "就", "都", "而", "及", "與", "這", "其", "但", "並", "個", "我",
"你", "他", "她", "它", "們", "我們", "你們", "他們", "她們", "它們", "在", "有", "人",
"這個", "那個", "如果", "因為", "所以", "可以", "沒有", "很", "非常", "得", "著", "過", "為", "再",
"吧", "呢", "啊", "哪", "那", "麼", "什麼", "誰", "哪裡", "哪裡", "怎麼", "怎麼樣", "為什麼", "將"
]
import shutil
def clear_previous_outputs():
directories_to_clear = [
f"{MAIN_OUTPUT_PATH}/base_processing",
f"{MAIN_OUTPUT_PATH}/filtered_output",
f"{S3_MINHASH_BASE_PATH}/{DUMP}/signatures",
f"{S3_MINHASH_BASE_PATH}/{DUMP}/buckets",
f"{S3_MINHASH_BASE_PATH}/{DUMP}/remove_ids",
f"{S3_MINHASH_BASE_PATH}/{DUMP}/deduped_output",
f"{MAIN_OUTPUT_PATH}/logs",
]
for directory in directories_to_clear:
if os.path.exists(directory):
print(f"Removing: {directory}")
shutil.rmtree(directory)
os.makedirs(directory, exist_ok=True)
def read_character_mappings(file_path: str) -> Tuple[Set[str], Set[str]]:
traditional = set()
simplified = set()
with open(file_path, 'r', encoding='utf-8') as f:
for line in f:
chars = line.strip().split('\t')
if len(chars) >= 2:
trad, simps = chars[0], chars[1].split()
traditional.add(trad)
simplified.update(simps)
return traditional, simplified
def read_traditional_characters(file_path: str) -> Set[str]:
with open(file_path, 'r', encoding='utf-8') as f:
return set(f.read().strip())
def create_simplified_filter(simplified_only: Set[str]) -> RegexFilter:
simplified_pattern = '|'.join(re.escape(char) for char in simplified_only)
return RegexFilter(
regex_exp=simplified_pattern,
exclusion_writer=JsonlWriter(
f"{FILTERING_OUTPUT_PATH}/removed/simplified/{DUMP}",
output_filename="${rank}.jsonl.gz"
)
)
def initial_filtering_pipeline(simplified_filter: RegexFilter) -> Tuple[LocalPipelineExecutor, LocalPipelineExecutor, LocalPipelineExecutor]:
# Initial filtering pipeline - Part 1
initial_executor_part1 = LocalPipelineExecutor(
pipeline=[
ParquetReader("./input/hugg-init/out-1720278098.parquet"),
LanguageFilter(languages=["zh"], language_threshold=0.65),
simplified_filter,
ParquetWriter(f"{MAIN_OUTPUT_PATH}/temp_output_part1")
],
tasks=8,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/base_processing/{DUMP}/part1",
)
# Initial filtering pipeline - Part 2
initial_executor_part2 = LocalPipelineExecutor(
pipeline=[
ParquetReader(f"{MAIN_OUTPUT_PATH}/temp_output_part1"),
GopherQualityFilter(
min_doc_words=15,
max_doc_words=10000,
min_avg_word_length=1,
max_avg_word_length=4,
max_symbol_word_ratio=0.1,
max_bullet_lines_ratio=0.9,
max_ellipsis_lines_ratio=0.3,
min_stop_words=1,
stop_words=chinese_stop_words,
exclusion_writer=JsonlWriter(f"{FILTERING_OUTPUT_PATH}/removed/4_gopher_qual/{DUMP}"),
language="zh",
),
ParquetWriter(f"{MAIN_OUTPUT_PATH}/temp_output_part2")
],
tasks=8,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/base_processing/{DUMP}/part2",
)
# Initial filtering pipeline - Part 3
initial_executor_part3 = LocalPipelineExecutor(
pipeline=[
ParquetReader(f"{MAIN_OUTPUT_PATH}/temp_output_part2"),
C4QualityFilter(
split_paragraph=False,
remove_citations=False,
filter_no_terminal_punct=False,
min_num_sentences=-1,
min_words_per_line=-1,
max_word_length=-1,
filter_lorem_ipsum=False,
filter_javascript=True,
filter_curly_bracket=True,
filter_policy=True,
language="zh",
exclusion_writer=JsonlWriter(f"{FILTERING_OUTPUT_PATH}/removed/5_c4/{DUMP}"),
),
ParquetWriter(f"{MAIN_OUTPUT_PATH}/filtered_output")
],
tasks=8,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/base_processing/{DUMP}/part3",
)
return initial_executor_part1, initial_executor_part2, initial_executor_part3
def deduplication_pipeline() -> Tuple[LocalPipelineExecutor, LocalPipelineExecutor, LocalPipelineExecutor, LocalPipelineExecutor]:
stage1 = LocalPipelineExecutor(
pipeline=[
ParquetReader(f"{MAIN_OUTPUT_PATH}/filtered_output"),
MinhashDedupSignature(
output_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/signatures",
config=minhash_config,
language="zh",
),
],
tasks=8,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/minhash/signatures",
)
stage2 = LocalPipelineExecutor(
pipeline=[
MinhashDedupBuckets(
input_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/signatures",
output_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/buckets",
config=minhash_config,
),
],
tasks=minhash_config.num_buckets,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/minhash/buckets",
)
stage3 = LocalPipelineExecutor(
pipeline=[
MinhashDedupCluster(
input_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/buckets",
output_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/remove_ids",
config=minhash_config,
),
],
tasks=1,
workers=1,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/minhash/clustering",
)
stage4 = LocalPipelineExecutor(
pipeline=[
ParquetReader(f"{MAIN_OUTPUT_PATH}/filtered_output"),
TokensCounter(),
MinhashDedupFilter(input_folder=f"{S3_MINHASH_BASE_PATH}/{DUMP}/remove_ids"),
PIIFormatter(),
ParquetWriter(f"{S3_MINHASH_BASE_PATH}/{DUMP}/deduped_output"),
],
tasks=8,
workers=8,
logging_dir=f"{MAIN_OUTPUT_PATH}/logs/minhash/filtering",
)
return stage1, stage2, stage3, stage4
def analyze_removed_documents(removed_dir: str, simplified_pattern: str, output_file: str):
removed_files = Path(removed_dir).glob('*.jsonl.gz')
simplified_regex = re.compile(f'[{simplified_pattern}]')
total_documents = 0
documents_with_simplified = 0
with gzip.open(output_file, 'wt', encoding='utf-8') as out_f:
for file in removed_files:
logger.info(f"Analyzing file: {file.name}")
with gzip.open(file, 'rt', encoding='utf-8') as in_f:
for line in in_f:
total_documents += 1
document = json.loads(line)
text = document.get('text', '')
matches = list(set(simplified_regex.findall(text)))
if matches:
documents_with_simplified += 1
result = {
"document_number": total_documents,
"simplified_characters": matches,
"text_snippet": text
}
out_f.write(json.dumps(result, ensure_ascii=False) + '\n')
logger.info(f"Analysis complete. Results written to {output_file}")
logger.info(f"Total documents analyzed: {total_documents}")
logger.info(f"Documents with simplified characters: {documents_with_simplified}")
def main(restart=False):
if restart:
clear_previous_outputs()
# Read character mappings
ts_traditional, ts_simplified = read_character_mappings("./input/ts-map/TSCharacters.txt")
st_traditional, st_simplified = read_character_mappings("./input/ts-map/STCharacters.txt")
# Combine sets
traditional = ts_traditional.union(st_traditional)
simplified = ts_simplified.union(st_simplified)
# Additional processing
white_list = set("床峰群秘霉庄痴雇简体踪")
additional_set = read_traditional_characters("./input/dict-tradi2/trad.txt")
print(f"Number of additional traditional characters: {len(additional_set)}")
simplified_only = simplified - traditional - white_list
# Create simplified filter
simplified_filter = create_simplified_filter(simplified_only)
# Run initial filtering pipeline
initial_executor_part1, initial_executor_part2, initial_executor_part3 = initial_filtering_pipeline(simplified_filter)
initial_executor_part1.run()
initial_executor_part2.run()
initial_executor_part3.run()
logger.info("Initial filtering complete. Starting analysis of removed documents.")
removed_dir = f"{FILTERING_OUTPUT_PATH}/removed/simplified/{DUMP}"
output_file = f"{MAIN_OUTPUT_PATH}/simplified_analysis_results.jsonl.gz"
# Create the simplified pattern
simplified_pattern = ''.join(simplified_only)
analyze_removed_documents(removed_dir, simplified_pattern, output_file)
# Run deduplication pipeline
# stage1, stage2, stage3, stage4 = deduplication_pipeline()
# stage1.run()
# stage2.run()
# stage3.run()
# stage4.run()
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(description="Run Chinese text processing pipeline")
parser.add_argument("--restart", action="store_true", help="Clear previous outputs and restart processing")
args = parser.parse_args()
main(restart=args.restart) |