Spaces:
Sleeping
Sleeping
File size: 12,371 Bytes
c0f31c1 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 | """
ParagraphChunker.py
A module for paragraph-level document chunking with token counting and preprocessing.
Features:
- Paragraph-based document splitting
- Content validation
- Multi-level delimiter detection
- Smart paragraph boundary detection
"""
import logging
import spacy
from typing import List, Optional
from pathlib import Path
from datetime import datetime
from langchain_core.documents import Document
from core.BaseChunker import BaseChunker
logger = logging.getLogger(__name__)
class ParagraphChunker(BaseChunker):
"""Handles document chunking at the paragraph level with token counting."""
PARAGRAPH_MIN_LENGTH = 50 # Minimum characters for a valid paragraph
def __init__(self, model_name=None, embedding_model=None):
"""
Initialize paragraph chunker with specified models.
Args:
model_name: Name of the model for tokenization
embedding_model: Model for generating embeddings
"""
super().__init__(model_name, embedding_model)
self.page_stats = []
# Initialize spaCy for NLP tasks
try:
self.nlp = spacy.load("en_core_web_sm")
except Exception as e:
logger.error(f"Error loading spaCy model: {e}")
import subprocess
logger.info("Installing spaCy model...")
subprocess.run(["python", "-m", "spacy", "download", "en_core_web_sm"],
capture_output=True)
self.nlp = spacy.load("en_core_web_sm")
def _split_into_paragraphs(self, text: str) -> List[str]:
"""
Split text into paragraphs using length and punctuation heuristics.
Args:
text: The text content to split
Returns:
List of paragraphs
"""
# Pre-clean the text
text = text.replace('\r', '\n')
# First, try double line breaks
paragraphs = text.split('\n\n')
# If that fails (PDF extraction issue), use sentence-based reconstruction
if len(paragraphs) <= 3:
print(f"PDF extraction flattened structure. Reconstructing from sentences...")
# Use spaCy for sentence detection
doc = self.nlp(text)
paragraphs = []
current_para = []
current_length = 0
for sent in doc.sents:
sent_text = sent.text.strip()
if not sent_text:
continue
# Add sentence to current paragraph
current_para.append(sent_text)
current_length += len(sent_text)
# Check if we should end the current paragraph
should_end_paragraph = (
# Paragraph is getting long (300-600 chars is typical)
current_length > 300 and
# Current sentence ends with proper punctuation
sent_text.endswith(('.', '!', '?')) and
# We have substantial content
len(current_para) >= 2
)
if should_end_paragraph:
paragraphs.append(' '.join(current_para))
current_para = []
current_length = 0
# Add the last paragraph
if current_para:
paragraphs.append(' '.join(current_para))
print(f"Reconstructed {len(paragraphs)} paragraphs using length heuristics")
# Clean and filter paragraphs
cleaned_paragraphs = []
for para in paragraphs:
clean_para = ' '.join(para.split())
if len(clean_para) >= self.PARAGRAPH_MIN_LENGTH:
cleaned_paragraphs.append(clean_para)
print(f"Final paragraph count: {len(cleaned_paragraphs)}")
return cleaned_paragraphs
def _process_single_paragraph(self, content: str, page_number: int,
para_number: int, preprocess: bool) -> Optional[Document]:
"""
Process a single paragraph with analysis and metadata.
Args:
content: The paragraph content
page_number: The page number
para_number: The paragraph number
preprocess: Whether to preprocess the text
Returns:
Document object with processed content and metadata, or None if paragraph is invalid
"""
# First check character length
if len(content.strip()) < self.PARAGRAPH_MIN_LENGTH:
self.page_stats.append(f"Paragraph {para_number} on page {page_number} is too short.")
return None
# Optionally preprocess the text
if preprocess:
content = self.preprocess_text(content)
# Analyze the paragraph and generate metadata
stats = self.analyze_text(content)
# Check token threshold
if stats["token_count"] < self.TOKEN_THRESHOLD:
self.page_stats.append(
f"Paragraph {para_number} on page {page_number} dropped: "
f"only {stats['token_count']} tokens"
)
return None
metadata = {
"page": page_number,
"paragraph": para_number,
"char_count": stats["char_count"],
"token_count": stats["token_count"],
"sentence_count": stats["sentence_count"],
"word_count": stats["word_count"],
"has_ocr": str(stats.get("has_content", True))
}
return Document(page_content=content, metadata=metadata)
def paragraph_process_document(self, file_path: str, preprocess: bool = False) -> List[Document]:
"""
Process PDF document paragraph by paragraph with analysis.
Args:
file_path: Path to the PDF file
preprocess: Whether to preprocess paragraph text
Returns:
List of Document objects, one per valid paragraph
"""
try:
self.page_stats = [] # Reset stats for this document
raw_pages = self.load_document(file_path)
processed_paragraphs = []
logger.info(f"Processing document with {len(raw_pages)} pages")
for page_idx, page in enumerate(raw_pages):
paragraphs = self._split_into_paragraphs(page.page_content)
logger.info(f"Page {page_idx+1}: Found {len(paragraphs)} paragraphs")
for para_idx, paragraph in enumerate(paragraphs):
processed_para = self._process_single_paragraph(
paragraph,
page_idx + 1,
para_idx + 1,
preprocess
)
if processed_para:
processed_paragraphs.append(processed_para)
# Output skipped paragraphs for transparency
if self.page_stats:
logger.info("\n".join(self.page_stats))
logger.info(f"Processed {len(processed_paragraphs)} valid paragraphs")
return processed_paragraphs
except Exception as e:
logger.error(f"Error in paragraph_process_document: {e}")
raise
def process_document(self, file_path: str, preprocess: bool = True) -> List[Document]:
"""
Process document using paragraph chunking strategy (implements abstract method).
Args:
file_path: Path to the PDF file
preprocess: Whether to preprocess paragraph text
Returns:
List of Document objects, one per valid paragraph
"""
return self.paragraph_process_document(file_path, preprocess)
def process_text_file(self, file_path: str, preprocess: bool = False) -> List[Document]:
"""
Process text file directly, preserving paragraph structure.
Args:
file_path: Path to the text file
preprocess: Whether to preprocess paragraph text
Returns:
List of Document objects, one per valid paragraph
"""
try:
# Load the text file directly
content = self.load_text_file(file_path)
# Clean the text using the same logic as PDF conversion
content = self.clean_text_for_processing(content)
# Split into paragraphs using double line breaks
paragraphs = content.split('\n\n')
logger.info(f"Found {len(paragraphs)} paragraphs in text file: {file_path}")
processed_paragraphs = []
file_name = Path(file_path).name
for para_idx, paragraph in enumerate(paragraphs):
paragraph = paragraph.strip()
if paragraph:
processed_para = self._process_single_paragraph_from_text(
paragraph,
file_path,
file_name,
para_idx + 1,
preprocess
)
if processed_para:
processed_paragraphs.append(processed_para)
logger.info(f"Processed {len(processed_paragraphs)} valid paragraphs from text file")
return processed_paragraphs
except Exception as e:
logger.error(f"Error processing text file: {e}")
raise
def _process_single_paragraph_from_text(self, content: str, file_path: str,
file_name: str, para_number: int,
preprocess: bool) -> Optional[Document]:
"""
Process a single paragraph from text file with analysis and metadata.
Args:
content: The paragraph content
file_path: Full path to the source file
file_name: Name of the source file
para_number: The paragraph number
preprocess: Whether to preprocess the text
Returns:
Document object with processed content and metadata, or None if paragraph is invalid
"""
# First check character length
if len(content.strip()) < self.PARAGRAPH_MIN_LENGTH:
logger.debug(f"Paragraph {para_number} too short ({len(content)} chars), skipping")
return None
# Preprocess if requested
if preprocess:
content = self.preprocess_text(content, remove_headers_footers=False)
# Analyze the paragraph
analysis = self.analyze_text(content)
# Validate content quality
if not self.is_content_valid(content):
logger.debug(f"Paragraph {para_number} failed content validation, skipping")
return None
# Create metadata
metadata = {
"source": file_path,
"file_name": file_name,
"file_type": "txt",
"paragraph": para_number,
"char_count": analysis["char_count"],
"token_count": analysis["token_count"],
"sentence_count": analysis["sentence_count"],
"word_count": analysis["word_count"],
"chunk_type": "paragraph",
"processing_timestamp": datetime.now().isoformat(),
}
# Create and return document
doc = Document(page_content=content, metadata=metadata)
logger.debug(f"Created paragraph {para_number}: {analysis['char_count']} chars, {analysis['token_count']} tokens")
return doc |