Spaces:
Running
Running
File size: 10,354 Bytes
94e74f0 |
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 |
# Standard library imports
import re
import logging
from difflib import SequenceMatcher
from typing import Tuple, Dict, Any, List, Optional
# Configure logging
logging.basicConfig(level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
def detect_duplicate_text_issues(text: str) -> Tuple[bool, Dict[str, Any]]:
"""
Detect if OCR text has duplication issues often found in handwritten document OCR
Args:
text: OCR text to analyze
Returns:
Tuple of (has_duplication_issues, details_dict)
"""
# Early exit for empty text
if not text or len(text) < 100:
return False, {"duplication_rate": 0.0, "details": "Text too short for analysis"}
# Look for repeated line patterns
lines = text.split('\n')
line_count = len(lines)
# Basic metrics
repeated_lines = 0
duplicate_sections = []
line_repetition_indices = []
# Check for exact line repetitions
seen_lines = {}
for i, line in enumerate(lines):
# Skip very short lines or empty lines
stripped = line.strip()
if len(stripped) < 5:
continue
if stripped in seen_lines:
repeated_lines += 1
line_repetition_indices.append((seen_lines[stripped], i))
else:
seen_lines[stripped] = i
# Calculate line repetition rate
line_repetition_rate = repeated_lines / max(1, line_count)
# Look for longer repeated sections using sequence matcher
text_blocks = [text[i:i+100] for i in range(0, len(text), 100) if i+100 <= len(text)]
block_count = len(text_blocks)
repeated_blocks = 0
for i in range(block_count):
for j in range(i+1, min(i+10, block_count)): # Only check nearby blocks for efficiency
matcher = SequenceMatcher(None, text_blocks[i], text_blocks[j])
similarity = matcher.ratio()
if similarity > 0.8: # High similarity threshold
repeated_blocks += 1
duplicate_sections.append((i, j, similarity))
break
# Calculate block repetition rate
block_repetition_rate = repeated_blocks / max(1, block_count)
# Combine metrics for overall duplication rate
duplication_rate = max(line_repetition_rate, block_repetition_rate)
# Detect patterns of repeated words in sequence (common OCR mistake)
word_pattern = r'\b(\w+)\s+\1\b'
repeated_words = len(re.findall(word_pattern, text))
repeated_words_rate = repeated_words / max(1, len(text.split()))
# Update duplication rate with word repetition
duplication_rate = max(duplication_rate, repeated_words_rate)
# Log detailed analysis
logger.info(f"OCR duplication analysis: line_repetition={line_repetition_rate:.2f}, "
f"block_repetition={block_repetition_rate:.2f}, "
f"word_repetition={repeated_words_rate:.2f}, "
f"final_rate={duplication_rate:.2f}")
# Determine if this is a serious issue
has_duplication = duplication_rate > 0.1
# Return detailed results
return has_duplication, {
"duplication_rate": duplication_rate,
"line_repetition_rate": line_repetition_rate,
"block_repetition_rate": block_repetition_rate,
"word_repetition_rate": repeated_words_rate,
"repeated_lines": repeated_lines,
"repeated_blocks": repeated_blocks,
"repeated_words": repeated_words,
"duplicate_sections": duplicate_sections[:10], # Only include the first 10 for brevity
"repetition_indices": line_repetition_indices[:10]
}
def get_enhanced_preprocessing_options(current_options: Optional[Dict[str, Any]] = None) -> Dict[str, Any]:
"""
Generate enhanced preprocessing options for improved OCR on handwritten documents
Args:
current_options: Current preprocessing options (if available)
Returns:
Dict of enhanced options
"""
# Start with current options or empty dict
options = current_options.copy() if current_options else {}
# Set document type to handwritten
options["document_type"] = "handwritten"
# Enhanced contrast - higher than normal for better handwriting extraction
options["contrast"] = 1.4 # Higher than default
# Apply grayscale
options["grayscale"] = True
# Apply adaptive thresholding optimized for handwriting
options["adaptive_threshold"] = True
options["threshold_block_size"] = 25 # Larger block size for handwriting
options["threshold_c"] = 10 # Adjusted C value for better handwriting detection
# Disable standard binarization which often loses handwriting detail
options["binarize"] = False
# Despeckle to reduce noise
options["denoise"] = True
# Enable handwriting-specific preprocessing
options["handwriting_mode"] = True
# Disable anything that might harm handwriting recognition
if "sharpen" in options:
options["sharpen"] = False
logger.info(f"Enhanced handwriting preprocessing options generated: {options}")
return options
def get_handwritten_specific_prompt(current_prompt: Optional[str] = None) -> str:
"""
Generate a specialized prompt for handwritten document OCR
Args:
current_prompt: Current prompt (if available)
Returns:
str: Enhanced prompt for handwritten documents
"""
# Base prompt for all handwritten documents
base_prompt = ("This is a handwritten document that requires careful transcription. "
"Please transcribe all visible handwritten text, preserving the original "
"line breaks, paragraph structure, and any special formatting or indentation. "
"Pay special attention to:\n"
"1. Words that may be difficult to read due to handwriting style\n"
"2. Any crossed-out text (indicate with [crossed out: possible text])\n"
"3. Insertions or annotations between lines or in margins\n"
"4. Maintain the spatial layout of the text as much as possible\n"
"5. If there are multiple columns or non-linear text, preserve the reading order\n\n"
"If you cannot read a word with confidence, indicate with [?] or provide your best guess as [word?].")
# If there's an existing prompt, combine them, otherwise just use the base
if current_prompt:
# Remove any redundant instructions about handwriting
lower_prompt = current_prompt.lower()
if "handwritten" in lower_prompt or "handwriting" in lower_prompt:
# Extract any unique instructions from the current prompt
# This logic is simplified and might need improvement
current_sentences = [s.strip() for s in current_prompt.split('.') if s.strip()]
handwriting_sentences = [s for s in current_sentences
if "handwritten" not in s.lower()
and "handwriting" not in s.lower()]
# Add unique instructions to our base prompt
if handwriting_sentences:
combined_prompt = base_prompt + "\n\nAdditional instructions:\n"
combined_prompt += ". ".join(handwriting_sentences) + "."
return combined_prompt
else:
# If no handwriting instructions in the current prompt, just append it
return f"{base_prompt}\n\nAdditional context from user:\n{current_prompt}"
return base_prompt
def clean_duplicated_text(text: str) -> str:
"""
Clean up duplicated text often found in OCR output for handwritten documents
Args:
text: OCR text to clean
Returns:
str: Cleaned text with duplications removed
"""
if not text:
return text
# Split into lines for line-based deduplication
lines = text.split('\n')
# Remove consecutive duplicate lines
deduped_lines = []
prev_line = None
for line in lines:
stripped = line.strip()
# Skip empty lines
if not stripped:
if not deduped_lines or deduped_lines[-1].strip():
deduped_lines.append(line) # Keep the first empty line
continue
# Skip if this line is a duplicate of the previous line
if stripped == prev_line:
continue
deduped_lines.append(line)
prev_line = stripped
# Re-join the deduplicated lines
deduped_text = '\n'.join(deduped_lines)
# Remove repeated words
word_pattern = r'\b(\w+)\s+\1\b'
deduped_text = re.sub(word_pattern, r'\1', deduped_text)
# Remove repeated phrases (3+ words)
# This is a simplified approach and might need improvement
words = deduped_text.split()
cleaned_words = []
i = 0
while i < len(words):
# Check for phrase repetition (phrases of 3 to 6 words)
found_repeat = False
for phrase_len in range(3, min(7, len(words) - i)):
phrase = ' '.join(words[i:i+phrase_len])
next_pos = i + phrase_len
if next_pos + phrase_len <= len(words):
next_phrase = ' '.join(words[next_pos:next_pos+phrase_len])
if phrase.lower() == next_phrase.lower():
# Found a repeated phrase, skip the second occurrence
cleaned_words.extend(words[i:i+phrase_len])
i = next_pos + phrase_len
found_repeat = True
break
if not found_repeat:
cleaned_words.append(words[i])
i += 1
# Rejoin the cleaned words
final_text = ' '.join(cleaned_words)
# Log the cleaning results
original_len = len(text)
cleaned_len = len(final_text)
reduction = 100 * (original_len - cleaned_len) / max(1, original_len)
logger.info(f"Text cleaning: removed {original_len - cleaned_len} chars ({reduction:.1f}% reduction)")
return final_text
|