Delete app2.py
Browse files
app2.py
DELETED
|
@@ -1,1707 +0,0 @@
|
|
| 1 |
-
#!/usr/bin/env python3
|
| 2 |
-
# -*- coding: utf-8 -*-
|
| 3 |
-
"""
|
| 4 |
-
🚀 Enhanced Bilingual Data Anonymization Benchmark System
|
| 5 |
-
====================================================================
|
| 6 |
-
نسخه پیشرفته با مقایسه واقعی متون اصلی و ناشناسسازی شده
|
| 7 |
-
"""
|
| 8 |
-
|
| 9 |
-
import gradio as gr
|
| 10 |
-
import pandas as pd
|
| 11 |
-
import numpy as np
|
| 12 |
-
import json
|
| 13 |
-
import time
|
| 14 |
-
import os
|
| 15 |
-
import re
|
| 16 |
-
import logging
|
| 17 |
-
import requests
|
| 18 |
-
from datetime import datetime
|
| 19 |
-
from functools import lru_cache
|
| 20 |
-
from packaging import version
|
| 21 |
-
from typing import Dict, List, Tuple, Any, Optional
|
| 22 |
-
import warnings
|
| 23 |
-
import gc
|
| 24 |
-
import threading
|
| 25 |
-
from collections import defaultdict, Counter
|
| 26 |
-
import hashlib
|
| 27 |
-
from concurrent.futures import ThreadPoolExecutor, as_completed
|
| 28 |
-
import multiprocessing
|
| 29 |
-
from dataclasses import dataclass
|
| 30 |
-
from difflib import SequenceMatcher
|
| 31 |
-
|
| 32 |
-
# Enhanced metrics imports
|
| 33 |
-
try:
|
| 34 |
-
import psutil
|
| 35 |
-
PSUTIL_AVAILABLE = True
|
| 36 |
-
except ImportError:
|
| 37 |
-
PSUTIL_AVAILABLE = False
|
| 38 |
-
|
| 39 |
-
try:
|
| 40 |
-
from sklearn.metrics import precision_score, recall_score, f1_score, accuracy_score, classification_report
|
| 41 |
-
from sklearn.feature_extraction.text import TfidfVectorizer
|
| 42 |
-
from sklearn.cluster import KMeans
|
| 43 |
-
SKLEARN_AVAILABLE = True
|
| 44 |
-
except ImportError:
|
| 45 |
-
SKLEARN_AVAILABLE = False
|
| 46 |
-
|
| 47 |
-
try:
|
| 48 |
-
import spacy
|
| 49 |
-
SPACY_AVAILABLE = True
|
| 50 |
-
except ImportError:
|
| 51 |
-
SPACY_AVAILABLE = False
|
| 52 |
-
|
| 53 |
-
warnings.filterwarnings('ignore')
|
| 54 |
-
|
| 55 |
-
# تنظیم logging
|
| 56 |
-
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
| 57 |
-
logger = logging.getLogger(__name__)
|
| 58 |
-
|
| 59 |
-
# =============================================================================
|
| 60 |
-
# Data Classes for Better Structure
|
| 61 |
-
# =============================================================================
|
| 62 |
-
|
| 63 |
-
@dataclass
|
| 64 |
-
class ComparisonResult:
|
| 65 |
-
"""نتیجه مقایسه متن اصلی با ناشناسسازی شده"""
|
| 66 |
-
index: int
|
| 67 |
-
success: bool
|
| 68 |
-
processing_time_ms: float
|
| 69 |
-
original_text: str
|
| 70 |
-
anonymized_text: str
|
| 71 |
-
original_length: int
|
| 72 |
-
anonymized_length: int
|
| 73 |
-
entities_should_anonymize: int
|
| 74 |
-
entities_correctly_anonymized: int
|
| 75 |
-
entities_missed: int
|
| 76 |
-
missed_entities_list: List[Dict]
|
| 77 |
-
anonymization_accuracy: float
|
| 78 |
-
precision: float
|
| 79 |
-
recall: float
|
| 80 |
-
f1_score: float
|
| 81 |
-
detected_language: str
|
| 82 |
-
confidence_score: float
|
| 83 |
-
memory_used_mb: float
|
| 84 |
-
entity_categories: Dict[str, int]
|
| 85 |
-
error: Optional[str] = None
|
| 86 |
-
|
| 87 |
-
@dataclass
|
| 88 |
-
class BenchmarkConfig:
|
| 89 |
-
"""تنظیمات بنچمارک"""
|
| 90 |
-
sample_size: int = 200
|
| 91 |
-
max_workers: int = 4
|
| 92 |
-
enable_parallel_processing: bool = True
|
| 93 |
-
enable_memory_profiling: bool = True
|
| 94 |
-
enable_language_detection: bool = True
|
| 95 |
-
enable_confidence_scoring: bool = True
|
| 96 |
-
stress_test_iterations: int = 50
|
| 97 |
-
enable_clustering_analysis: bool = False
|
| 98 |
-
|
| 99 |
-
# =============================================================================
|
| 100 |
-
# Enhanced Pattern Library
|
| 101 |
-
# =============================================================================
|
| 102 |
-
|
| 103 |
-
class EnhancedPatternLibrary:
|
| 104 |
-
"""کتابخانه الگوهای پیشرفته"""
|
| 105 |
-
|
| 106 |
-
def __init__(self):
|
| 107 |
-
self.patterns = self._load_enhanced_patterns()
|
| 108 |
-
self.compiled_patterns = self._compile_patterns()
|
| 109 |
-
|
| 110 |
-
def _load_enhanced_patterns(self):
|
| 111 |
-
"""بارگذاری الگوهای پیشرفته و جامعتر"""
|
| 112 |
-
return {
|
| 113 |
-
'PERSIAN_PERSON': [
|
| 114 |
-
r'آقای\s+([آ-ی\u200C]{2,}(?:\s+[آ-ی\u200C]{2,})*)',
|
| 115 |
-
r'خانم\s+([آ-ی\u200C]{2,}(?:\s+[آ-ی\u200C]{2,})*)',
|
| 116 |
-
r'مهندس\s+([آ-ی\u200C]{2,}(?:\s+[آ-ی\u200C]{2,})*)',
|
| 117 |
-
r'دکتر\s+([آ-ی\u200C]{2,}(?:\s+[آ-ی\u200C]{2,})*)',
|
| 118 |
-
r'استاد\s+([آ-ی\u200C]{2,}(?:\s+[آ-ی\u200C]{2,})*)',
|
| 119 |
-
r'([آ-ی\u200C]{3,}\s+[آ-ی\u200C]{3,})(?=\s+مدیرعامل|\s+رئیس|\s+مدیر|[،.]|\s*$)',
|
| 120 |
-
r'\b([آ-ی\u200C]{3,}(?:\s+[آ-ی\u200C]{3,}){1,2})\b',
|
| 121 |
-
# نامهای متداول فارسی
|
| 122 |
-
r'\b(احمد|علی|حسن|حسین|محمد|رضا|مهدی|امیر|سعید|مجید|فرهاد|بهرام|کامران|داود|یوسف|ابراهیم)\s+[آ-ی\u200C]{3,}\b',
|
| 123 |
-
r'\b[آ-ی\u200C]{3,}\s+(احمدی|علوی|حسینی|محمدی|رضایی|کریمی|موسوی|صادقی|مرادی|فرهادی)\b',
|
| 124 |
-
],
|
| 125 |
-
'ENGLISH_PERSON': [
|
| 126 |
-
r'(Mr\.|Mrs\.|Ms\.|Dr\.|Prof\.)\s+([A-Z][a-z]{2,}(?:\s+[A-Z][a-z]{2,})*)',
|
| 127 |
-
r'\b([A-Z][a-z]{2,}\s+[A-Z][a-z]{2,})(?=\s+(?:CEO|President|Manager|Director|said|stated|announced))',
|
| 128 |
-
r'\b([A-Z][a-z]{2,}(?:\s+[A-Z]\.)*\s+[A-Z][a-z]{2,})\b',
|
| 129 |
-
# نامهای متداول انگلیسی
|
| 130 |
-
r'\b(John|James|Michael|William|David|Richard|Joseph|Thomas|Christopher|Daniel|Paul|Mark|Donald|Steven|Andrew|Kenneth|Paul|Joshua|Kevin|Brian|George|Timothy|Ronald|Jason|Edward|Jeffrey|Ryan|Jacob|Gary|Nicholas|Eric|Jonathan|Stephen|Larry|Justin|Scott|Brandon|Benjamin|Samuel|Gregory|Frank|Raymond|Alexander|Patrick|Jack|Dennis|Jerry|Tyler|Aaron|Jose|Henry|Adam|Douglas|Nathan|Zachary|Kyle)\s+[A-Z][a-z]{2,}\b',
|
| 131 |
-
r'\b[A-Z][a-z]{2,}\s+(Smith|Johnson|Williams|Brown|Jones|Garcia|Miller|Davis|Rodriguez|Martinez|Hernandez|Lopez|Gonzalez|Wilson|Anderson|Thomas|Taylor|Moore|Jackson|Martin|Lee|Perez|Thompson|White|Harris|Sanchez|Clark|Ramirez|Lewis|Robinson|Walker|Young|Allen|King|Wright|Scott|Torres|Nguyen|Hill|Flores|Green|Adams|Nelson|Baker|Hall|Rivera|Campbell|Mitchell|Carter|Roberts)\b',
|
| 132 |
-
],
|
| 133 |
-
'ENHANCED_PHONE': [
|
| 134 |
-
r'(?:تلفن|موبایل|تماس)[\s:]*(?:\+98|0098)?(?:0)?([۰-۹0-9]{10,11})',
|
| 135 |
-
r'(?:\+98|0098)[\s\-]?([۰-۹0-9]{2,3})[\s\-]?([۰-۹0-9]{7,8})',
|
| 136 |
-
r'(?:^|[^\d])0([۰-۹0-9]{2,3})[\s\-]?([۰-۹0-9]{7,8})',
|
| 137 |
-
r'\b([۰-۹0-9]{4})[\s\-]([۰-۹0-9]{3})[\s\-]([۰-۹0-9]{4})\b',
|
| 138 |
-
r'\b(\+\d{1,3}[\s\-]?\d{3}[\s\-]?\d{3}[\s\-]?\d{4})\b',
|
| 139 |
-
r'\b(\d{3}[-.]?\d{3}[-.]?\d{4})\b',
|
| 140 |
-
r'\b([۰-۹]{4}[\s\-][۰-۹]{7})\b',
|
| 141 |
-
r'\b(09[۰-۹0-9]{9})\b', # شماره موبایل ایرانی
|
| 142 |
-
# الگوهای تلفن بینالمللی
|
| 143 |
-
r'\b(\+1[\s\-]?\d{3}[\s\-]?\d{3}[\s\-]?\d{4})\b', # US
|
| 144 |
-
r'\b(\+44[\s\-]?\d{4}[\s\-]?\d{6})\b', # UK
|
| 145 |
-
],
|
| 146 |
-
'ENHANCED_EMAIL': [
|
| 147 |
-
r'\b([a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,})\b',
|
| 148 |
-
r'(?:ایمیل|email)[\s:]*([a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,})',
|
| 149 |
-
r'\b([a-zA-Z0-9._%+-]+@(?:gmail|yahoo|hotmail|outlook|aol)\.com)\b',
|
| 150 |
-
],
|
| 151 |
-
'ENHANCED_NATIONAL_ID': [
|
| 152 |
-
r'(?:کد\s*ملی|شناسه\s*ملی)[\s:]*([۰-۹0-9]{10})',
|
| 153 |
-
r'(?:National\s*ID)[\s:]*([0-9]{10})',
|
| 154 |
-
r'(?:شماره\s*شناسنامه)[\s:]*([۰-۹0-9]{1,10})',
|
| 155 |
-
r'\b([۰-۹0-9]{10})\b',
|
| 156 |
-
r'(?:SSN|Social\s*Security)[\s:]*([0-9]{3}-[0-9]{2}-[0-9]{4})',
|
| 157 |
-
],
|
| 158 |
-
'ENHANCED_BANK_ACCOUNT': [
|
| 159 |
-
r'(?:شماره\s*حساب|حساب\s*بانکی)[\s:]*([۰-۹0-9\-]{10,20})',
|
| 160 |
-
r'(?:شبا|IBAN)[\s:]*IR([۰-۹0-9]{24})',
|
| 161 |
-
r'(?:کارت\s*بانکی)[\s:]*([۰-۹0-9]{4}[\s\-]?[۰-۹0-9]{4}[\s\-]?[۰-۹0-9]{4}[\s\-]?[۰-۹0-9]{4})',
|
| 162 |
-
r'\b([0-9]{4}[\s\-]?[0-9]{4}[\s\-]?[0-9]{4}[\s\-]?[0-9]{4})\b',
|
| 163 |
-
r'\b([0-9]{10,20})\b(?=.*(?:account|حساب))',
|
| 164 |
-
# الگوهای کارت اعتباری
|
| 165 |
-
r'\b(4[0-9]{12}(?:[0-9]{3})?)\b', # Visa
|
| 166 |
-
r'\b(5[1-5][0-9]{14})\b', # MasterCard
|
| 167 |
-
r'\b(3[47][0-9]{13})\b', # American Express
|
| 168 |
-
],
|
| 169 |
-
'ENHANCED_AMOUNT': [
|
| 170 |
-
r'(?:مبلغ)?\s*([۰-۹0-9,]+)\s*(?:میلیون|میلیارد|هزار)?\s*(?:تومان|ریال)',
|
| 171 |
-
r'\$([0-9,]+(?:\.[0-9]{2})?)\s*(?:million|billion|thousand|M|B|K)?',
|
| 172 |
-
r'€([0-9,]+(?:\.[0-9]{2})?)',
|
| 173 |
-
r'£([0-9,]+(?:\.[0-9]{2})?)',
|
| 174 |
-
r'\b([0-9,]+(?:\.[0-9]{2})?)\s*(?:dollar|euro|pound|USD|EUR|GBP)s?\b',
|
| 175 |
-
r'\b([۰-۹0-9,]+)\s*(?:درهم|دینار|ین|یوان)\b',
|
| 176 |
-
],
|
| 177 |
-
'ENHANCED_DATE': [
|
| 178 |
-
r'([۰-۹0-9]{4})[/\-]([۰-۹0-9]{1,2})[/\-]([۰-۹0-9]{1,2})',
|
| 179 |
-
r'([۰-۹0-9]{1,2})[/\-]([۰-۹0-9]{1,2})[/\-]([۰-۹0-9]{4})',
|
| 180 |
-
r'(\d{1,2})\s+(January|February|March|April|May|June|July|August|September|October|November|December)\s+(\d{4})',
|
| 181 |
-
r'(\d{1,2})\s+(فروردین|اردیبهشت|خرداد|تیر|مرداد|شهریور|مهر|آبان|آذر|دی|بهمن|اسفند)\s+(\d{4})',
|
| 182 |
-
r'\b(\d{1,2}/\d{1,2}/\d{2,4})\b',
|
| 183 |
-
r'\b(\d{4}-\d{2}-\d{2})\b', # ISO format
|
| 184 |
-
r'\b([۰-۹]{4}/[۰-۹]{1,2}/[۰-۹]{1,2})\b',
|
| 185 |
-
],
|
| 186 |
-
'ENHANCED_COMPANY': [
|
| 187 |
-
r'(?:شرکت)\s+([آ-ی\u200C\s]{3,}?)(?=\s+در|\s+که|\s+با|[،.]|\s*$)',
|
| 188 |
-
r'(?:بانک)\s+([آ-ی\u200C\s]{3,})',
|
| 189 |
-
r'\b([A-Z][a-zA-Z\s&]{2,}(?:Inc|Corp|Corporation|Company|Ltd|Limited|LLC|LLP))\b',
|
| 190 |
-
r'\b([A-Z][a-zA-Z\s&]{2,}Bank)\b',
|
| 191 |
-
r'\b(Apple|Google|Microsoft|Amazon|Facebook|Tesla|Netflix|IBM|Oracle|Samsung|Sony|Toyota|BMW|Mercedes|Volkswagen|Ford|General Motors)\b',
|
| 192 |
-
r'\b([آ-ی\u200C]{3,}(?:\s+[آ-ی\u200C]{3,})*)\s+(?:شرکت|گروه|هلدینگ|صنایع)\b',
|
| 193 |
-
],
|
| 194 |
-
'ENHANCED_LOCATION': [
|
| 195 |
-
r'(?:شهر|استان)\s+([آ-ی\u200C\s]{2,})',
|
| 196 |
-
r'\b(تهران|اصفهان|شیراز|مشهد|تبریز|اهواز|کرج|قم|کرمان|یزد|ساری|گرگان|رشت|ارومیه|زاهدان|کرمانشاه|همدان|اراک|قزوین|زنجان|سنندج|ایلام|یاسوج|بجنورد|گر��ان|بندرعباس|بوشهر)\b',
|
| 197 |
-
r'(خیابان|کوچه|بلوار|میدان)\s+([آ-ی\u200C\s]{2,})',
|
| 198 |
-
r'پلاک\s*([۰-۹0-9]+)',
|
| 199 |
-
r'\b([A-Z][a-zA-Z\s]{2,}(?:Street|Avenue|Road|Boulevard|Drive|Lane|Way))\b',
|
| 200 |
-
r'\b([A-Z][a-zA-Z\s]{2,},\s*[A-Z]{2}\s+\d{5})\b',
|
| 201 |
-
r'\b(New York|Los Angeles|Chicago|Houston|Phoenix|Philadelphia|San Antonio|San Diego|Dallas|San Jose|Austin|Jacksonville|Fort Worth|Columbus|Charlotte|San Francisco|Indianapolis|Seattle|Denver|Washington|Boston|El Paso|Detroit|Nashville|Portland|Memphis|Oklahoma City|Las Vegas|Louisville|Baltimore|Milwaukee|Albuquerque|Tucson|Fresno|Sacramento|Long Beach|Kansas City|Mesa|Virginia Beach|Atlanta|Colorado Springs|Omaha|Raleigh|Miami|Oakland|Minneapolis|Tulsa|Cleveland|Wichita|Arlington)\b',
|
| 202 |
-
# کشورها
|
| 203 |
-
r'\b(Iran|Iraq|Turkey|Afghanistan|Pakistan|India|China|Russia|Germany|France|Italy|Spain|United Kingdom|Canada|Australia|Japan|South Korea|Brazil|Mexico|Argentina)\b',
|
| 204 |
-
r'\b(ایران|عراق|ترکیه|افغانستان|پاکستان|هندوستان|چین|روسیه|آلمان|فرانسه|ایتالیا|اسپانیا|انگلستان|کانادا|استرالیا|ژاپن|کره جنوبی|برزیل|مکزیک|آرژانتین)\b',
|
| 205 |
-
],
|
| 206 |
-
'ENHANCED_PERCENTAGE': [
|
| 207 |
-
r'([۰-۹0-9]+(?:\.[۰-۹0-9]+)?)\s*درصد',
|
| 208 |
-
r'([۰-۹0-9]+(?:\.[۰-۹0-9]+)?)\s*%',
|
| 209 |
-
r'(?:رشد|افزایش|کاهش|تغییر)\s+([۰-۹0-9]+(?:\.[۰-۹0-9]+)?)\s*درصدی',
|
| 210 |
-
r'\b([0-9]+(?:\.[0-9]+)?)\s*percent\b',
|
| 211 |
-
],
|
| 212 |
-
'ENHANCED_ADDRESS': [
|
| 213 |
-
r'([آ-ی\u200C\s]{2,})\s*،\s*(خیابان|کوچه|بلوار|میدان)\s+([آ-ی\u200C\s]{2,})\s*،\s*پلاک\s*([۰-۹0-9]+)',
|
| 214 |
-
r'([0-9]+)\s+([A-Z][a-zA-Z\s]{2,}(?:Street|Avenue|Road|Boulevard|Drive|Lane))',
|
| 215 |
-
r'\b([0-9]+\s+[A-Z][a-zA-Z\s]{2,},\s*[A-Z][a-zA-Z\s]{2,},\s*[A-Z]{2}\s+[0-9]{5})\b',
|
| 216 |
-
],
|
| 217 |
-
'ENHANCED_IP_ADDRESS': [
|
| 218 |
-
r'\b((?:[0-9]{1,3}\.){3}[0-9]{1,3})\b',
|
| 219 |
-
r'\b([0-9a-fA-F:]+::[0-9a-fA-F:]+)\b', # IPv6
|
| 220 |
-
],
|
| 221 |
-
'ENHANCED_URL': [
|
| 222 |
-
r'\b(https?://[^\s]+)\b',
|
| 223 |
-
r'\b(www\.[^\s]+\.[a-zA-Z]{2,})\b',
|
| 224 |
-
]
|
| 225 |
-
}
|
| 226 |
-
|
| 227 |
-
def _compile_patterns(self):
|
| 228 |
-
"""کامپایل الگوهای regex برای بهبود عملکرد"""
|
| 229 |
-
compiled = {}
|
| 230 |
-
for category, patterns in self.patterns.items():
|
| 231 |
-
compiled[category] = [re.compile(pattern, re.IGNORECASE | re.UNICODE) for pattern in patterns]
|
| 232 |
-
return compiled
|
| 233 |
-
|
| 234 |
-
def find_matches(self, text: str, category: Optional[str] = None) -> Dict[str, List[Tuple[str, int, int]]]:
|
| 235 |
-
"""پیدا کردن تطبیقها با بازگشت موقعیت"""
|
| 236 |
-
matches = defaultdict(list)
|
| 237 |
-
|
| 238 |
-
patterns_to_check = {category: self.compiled_patterns[category]} if category else self.compiled_patterns
|
| 239 |
-
|
| 240 |
-
for cat, compiled_patterns in patterns_to_check.items():
|
| 241 |
-
for pattern in compiled_patterns:
|
| 242 |
-
for match in pattern.finditer(text):
|
| 243 |
-
matched_text = match.group(1) if match.groups() else match.group(0)
|
| 244 |
-
if matched_text and matched_text.strip():
|
| 245 |
-
matches[cat].append((matched_text.strip(), match.start(), match.end()))
|
| 246 |
-
|
| 247 |
-
return dict(matches)
|
| 248 |
-
|
| 249 |
-
# =============================================================================
|
| 250 |
-
# Enhanced Comparison Anonymizer
|
| 251 |
-
# =============================================================================
|
| 252 |
-
|
| 253 |
-
class EnhancedComparisonAnonymizer:
|
| 254 |
-
"""سیستم مقایسه و ارزیابی ناشناسسازی"""
|
| 255 |
-
|
| 256 |
-
def __init__(self, config: BenchmarkConfig):
|
| 257 |
-
self.config = config
|
| 258 |
-
self.pattern_lib = EnhancedPatternLibrary()
|
| 259 |
-
self.processing_stats = {
|
| 260 |
-
'total_processed': 0,
|
| 261 |
-
'total_entities_should_anonymize': 0,
|
| 262 |
-
'total_entities_correctly_anonymized': 0,
|
| 263 |
-
'total_entities_missed': 0,
|
| 264 |
-
'language_distribution': Counter(),
|
| 265 |
-
}
|
| 266 |
-
|
| 267 |
-
# Load models if available
|
| 268 |
-
self.models_loaded = False
|
| 269 |
-
self.load_models()
|
| 270 |
-
|
| 271 |
-
def load_models(self):
|
| 272 |
-
"""بارگذاری مدلهای NER"""
|
| 273 |
-
try:
|
| 274 |
-
if SPACY_AVAILABLE:
|
| 275 |
-
try:
|
| 276 |
-
import spacy
|
| 277 |
-
self.nlp_en = spacy.load("en_core_web_sm")
|
| 278 |
-
logger.info("✅ English spaCy model loaded")
|
| 279 |
-
except OSError:
|
| 280 |
-
logger.warning("⚠️ English spaCy model not found")
|
| 281 |
-
self.nlp_en = None
|
| 282 |
-
|
| 283 |
-
try:
|
| 284 |
-
self.nlp_fa = spacy.load("fa_core_news_sm")
|
| 285 |
-
logger.info("✅ Persian spaCy model loaded")
|
| 286 |
-
except OSError:
|
| 287 |
-
logger.warning("⚠️ Persian spaCy model not found")
|
| 288 |
-
self.nlp_fa = None
|
| 289 |
-
|
| 290 |
-
self.models_loaded = (self.nlp_en is not None) or (self.nlp_fa is not None)
|
| 291 |
-
|
| 292 |
-
except Exception as e:
|
| 293 |
-
logger.error(f"❌ Error loading models: {e}")
|
| 294 |
-
self.models_loaded = False
|
| 295 |
-
|
| 296 |
-
def detect_language(self, text: str) -> str:
|
| 297 |
-
"""تشخیص زبان متن"""
|
| 298 |
-
persian_chars = len(re.findall(r'[\u0600-\u06FF]', text))
|
| 299 |
-
english_chars = len(re.findall(r'[a-zA-Z]', text))
|
| 300 |
-
total_chars = persian_chars + english_chars
|
| 301 |
-
|
| 302 |
-
if total_chars == 0:
|
| 303 |
-
return 'unknown'
|
| 304 |
-
|
| 305 |
-
persian_ratio = persian_chars / total_chars
|
| 306 |
-
english_ratio = english_chars / total_chars
|
| 307 |
-
|
| 308 |
-
if persian_ratio > 0.7:
|
| 309 |
-
return 'fa'
|
| 310 |
-
elif english_ratio > 0.7:
|
| 311 |
-
return 'en'
|
| 312 |
-
elif persian_ratio > 0.3 and english_ratio > 0.3:
|
| 313 |
-
return 'mixed'
|
| 314 |
-
elif persian_ratio > english_ratio:
|
| 315 |
-
return 'fa'
|
| 316 |
-
else:
|
| 317 |
-
return 'en'
|
| 318 |
-
|
| 319 |
-
def compare_texts(self, original_text: str, anonymized_text: str) -> ComparisonResult:
|
| 320 |
-
"""مقایسه متن اصلی با ناشناسسازی شده"""
|
| 321 |
-
start_time = time.time()
|
| 322 |
-
memory_before = self._get_memory_usage()
|
| 323 |
-
|
| 324 |
-
try:
|
| 325 |
-
# تشخیص زبان
|
| 326 |
-
detected_lang = self.detect_language(original_text)
|
| 327 |
-
|
| 328 |
-
# پیدا کردن موجودیتهایی که باید ناشناسسازی شوند
|
| 329 |
-
original_entities = self._find_all_entities(original_text, detected_lang)
|
| 330 |
-
|
| 331 |
-
# بررسی اینکه کدام موجودیتها ناشناسسازی شدهاند
|
| 332 |
-
correctly_anonymized, missed_entities = self._check_anonymization_quality(
|
| 333 |
-
original_text, anonymized_text, original_entities
|
| 334 |
-
)
|
| 335 |
-
|
| 336 |
-
# محاسبه متریکها
|
| 337 |
-
total_should_anonymize = len(original_entities)
|
| 338 |
-
correctly_anonymized_count = len(correctly_anonymized)
|
| 339 |
-
missed_count = len(missed_entities)
|
| 340 |
-
|
| 341 |
-
# محاسبه دقت، بازخوانی و F1
|
| 342 |
-
precision = correctly_anonymized_count / max(1, correctly_anonymized_count + self._count_false_positives(anonymized_text))
|
| 343 |
-
recall = correctly_anonymized_count / max(1, total_should_anonymize)
|
| 344 |
-
f1 = 2 * (precision * recall) / max(0.001, precision + recall)
|
| 345 |
-
anonymization_accuracy = correctly_anonymized_count / max(1, total_should_anonymize)
|
| 346 |
-
|
| 347 |
-
# محاسبه آمار
|
| 348 |
-
processing_time = (time.time() - start_time) * 1000
|
| 349 |
-
memory_after = self._get_memory_usage()
|
| 350 |
-
memory_used = max(0, memory_after - memory_before)
|
| 351 |
-
|
| 352 |
-
confidence = self._calculate_confidence_score(original_entities, correctly_anonymized_count, total_should_anonymize)
|
| 353 |
-
|
| 354 |
-
# شمارش دستهها
|
| 355 |
-
entity_categories = defaultdict(int)
|
| 356 |
-
for entity in original_entities:
|
| 357 |
-
entity_categories[entity['category']] += 1
|
| 358 |
-
|
| 359 |
-
# بهروزرسانی آمار کلی
|
| 360 |
-
self.processing_stats['total_processed'] += 1
|
| 361 |
-
self.processing_stats['total_entities_should_anonymize'] += total_should_anonymize
|
| 362 |
-
self.processing_stats['total_entities_correctly_anonymized'] += correctly_anonymized_count
|
| 363 |
-
self.processing_stats['total_entities_missed'] += missed_count
|
| 364 |
-
self.processing_stats['language_distribution'][detected_lang] += 1
|
| 365 |
-
|
| 366 |
-
return ComparisonResult(
|
| 367 |
-
index=self.processing_stats['total_processed'],
|
| 368 |
-
success=True,
|
| 369 |
-
processing_time_ms=processing_time,
|
| 370 |
-
original_text=original_text[:500] + "..." if len(original_text) > 500 else original_text,
|
| 371 |
-
anonymized_text=anonymized_text[:500] + "..." if len(anonymized_text) > 500 else anonymized_text,
|
| 372 |
-
original_length=len(original_text),
|
| 373 |
-
anonymized_length=len(anonymized_text),
|
| 374 |
-
entities_should_anonymize=total_should_anonymize,
|
| 375 |
-
entities_correctly_anonymized=correctly_anonymized_count,
|
| 376 |
-
entities_missed=missed_count,
|
| 377 |
-
missed_entities_list=missed_entities,
|
| 378 |
-
anonymization_accuracy=anonymization_accuracy,
|
| 379 |
-
precision=precision,
|
| 380 |
-
recall=recall,
|
| 381 |
-
f1_score=f1,
|
| 382 |
-
detected_language=detected_lang,
|
| 383 |
-
confidence_score=confidence,
|
| 384 |
-
memory_used_mb=memory_used,
|
| 385 |
-
entity_categories=dict(entity_categories)
|
| 386 |
-
)
|
| 387 |
-
|
| 388 |
-
except Exception as e:
|
| 389 |
-
processing_time = (time.time() - start_time) * 1000
|
| 390 |
-
return ComparisonResult(
|
| 391 |
-
index=self.processing_stats['total_processed'],
|
| 392 |
-
success=False,
|
| 393 |
-
processing_time_ms=processing_time,
|
| 394 |
-
original_text=original_text[:200] + "..." if len(original_text) > 200 else original_text,
|
| 395 |
-
anonymized_text=anonymized_text[:200] + "..." if len(anonymized_text) > 200 else anonymized_text,
|
| 396 |
-
original_length=len(original_text),
|
| 397 |
-
anonymized_length=len(anonymized_text),
|
| 398 |
-
entities_should_anonymize=0,
|
| 399 |
-
entities_correctly_anonymized=0,
|
| 400 |
-
entities_missed=0,
|
| 401 |
-
missed_entities_list=[],
|
| 402 |
-
anonymization_accuracy=0.0,
|
| 403 |
-
precision=0.0,
|
| 404 |
-
recall=0.0,
|
| 405 |
-
f1_score=0.0,
|
| 406 |
-
detected_language='unknown',
|
| 407 |
-
confidence_score=0.0,
|
| 408 |
-
memory_used_mb=0.0,
|
| 409 |
-
entity_categories={},
|
| 410 |
-
error=str(e)
|
| 411 |
-
)
|
| 412 |
-
|
| 413 |
-
def _find_all_entities(self, text: str, language: str) -> List[Dict]:
|
| 414 |
-
"""پیدا کردن تمام موجودیتهایی که باید ناشناسسازی شوند"""
|
| 415 |
-
entities = []
|
| 416 |
-
|
| 417 |
-
# استفاده از pattern matching
|
| 418 |
-
pattern_matches = self.pattern_lib.find_matches(text)
|
| 419 |
-
|
| 420 |
-
for category, matches in pattern_matches.items():
|
| 421 |
-
for match_text, start, end in matches:
|
| 422 |
-
entities.append({
|
| 423 |
-
'text': match_text,
|
| 424 |
-
'category': category,
|
| 425 |
-
'start': start,
|
| 426 |
-
'end': end,
|
| 427 |
-
'source': 'pattern'
|
| 428 |
-
})
|
| 429 |
-
|
| 430 |
-
# استفاده از NER اگر در دسترس باشد
|
| 431 |
-
if self.models_loaded:
|
| 432 |
-
ner_entities = self._extract_entities_with_ner(text, language)
|
| 433 |
-
for entity in ner_entities:
|
| 434 |
-
entities.append({
|
| 435 |
-
'text': entity['text'],
|
| 436 |
-
'category': self._map_ner_label(entity['label']),
|
| 437 |
-
'start': entity['start'],
|
| 438 |
-
'end': entity['end'],
|
| 439 |
-
'source': entity['source']
|
| 440 |
-
})
|
| 441 |
-
|
| 442 |
-
# حذف تداخلها
|
| 443 |
-
entities = self._remove_overlapping_entities(entities)
|
| 444 |
-
|
| 445 |
-
return entities
|
| 446 |
-
|
| 447 |
-
def _extract_entities_with_ner(self, text: str, language: str) -> List[Dict]:
|
| 448 |
-
"""استخراج entities با مدلهای NER"""
|
| 449 |
-
entities = []
|
| 450 |
-
|
| 451 |
-
try:
|
| 452 |
-
if language in ['en', 'mixed'] and hasattr(self, 'nlp_en') and self.nlp_en:
|
| 453 |
-
doc = self.nlp_en(text)
|
| 454 |
-
for ent in doc.ents:
|
| 455 |
-
entities.append({
|
| 456 |
-
'text': ent.text,
|
| 457 |
-
'label': ent.label_,
|
| 458 |
-
'start': ent.start_char,
|
| 459 |
-
'end': ent.end_char,
|
| 460 |
-
'source': 'spacy_en'
|
| 461 |
-
})
|
| 462 |
-
|
| 463 |
-
if language in ['fa', 'mixed'] and hasattr(self, 'nlp_fa') and self.nlp_fa:
|
| 464 |
-
doc = self.nlp_fa(text)
|
| 465 |
-
for ent in doc.ents:
|
| 466 |
-
entities.append({
|
| 467 |
-
'text': ent.text,
|
| 468 |
-
'label': ent.label_,
|
| 469 |
-
'start': ent.start_char,
|
| 470 |
-
'end': ent.end_char,
|
| 471 |
-
'source': 'spacy_fa'
|
| 472 |
-
})
|
| 473 |
-
|
| 474 |
-
except Exception as e:
|
| 475 |
-
logger.error(f"Error in NER extraction: {e}")
|
| 476 |
-
|
| 477 |
-
return entities
|
| 478 |
-
|
| 479 |
-
def _map_ner_label(self, ner_label: str) -> str:
|
| 480 |
-
"""نقشهبرداری برچسبهای NER"""
|
| 481 |
-
mapping = {
|
| 482 |
-
'PERSON': 'PERSIAN_PERSON',
|
| 483 |
-
'PER': 'PERSIAN_PERSON',
|
| 484 |
-
'ORG': 'ENHANCED_COMPANY',
|
| 485 |
-
'ORGANIZATION': 'ENHANCED_COMPANY',
|
| 486 |
-
'LOC': 'ENHANCED_LOCATION',
|
| 487 |
-
'LOCATION': 'ENHANCED_LOCATION',
|
| 488 |
-
'GPE': 'ENHANCED_LOCATION',
|
| 489 |
-
'MONEY': 'ENHANCED_AMOUNT',
|
| 490 |
-
'PERCENT': 'ENHANCED_PERCENTAGE',
|
| 491 |
-
'DATE': 'ENHANCED_DATE',
|
| 492 |
-
'TIME': 'ENHANCED_DATE'
|
| 493 |
-
}
|
| 494 |
-
return mapping.get(ner_label.upper(), 'OTHER')
|
| 495 |
-
|
| 496 |
-
def _remove_overlapping_entities(self, entities: List[Dict]) -> List[Dict]:
|
| 497 |
-
"""حذف موجودیتهای همپوشان"""
|
| 498 |
-
entities.sort(key=lambda x: (x['start'], x['end'] - x['start']))
|
| 499 |
-
|
| 500 |
-
filtered_entities = []
|
| 501 |
-
used_positions = []
|
| 502 |
-
|
| 503 |
-
for entity in entities:
|
| 504 |
-
start, end = entity['start'], entity['end']
|
| 505 |
-
overlaps = any(not (end <= pos_start or start >= pos_end) for pos_start, pos_end in used_positions)
|
| 506 |
-
|
| 507 |
-
if not overlaps:
|
| 508 |
-
filtered_entities.append(entity)
|
| 509 |
-
used_positions.append((start, end))
|
| 510 |
-
|
| 511 |
-
return filtered_entities
|
| 512 |
-
|
| 513 |
-
def _check_anonymization_quality(self, original_text: str, anonymized_text: str, entities: List[Dict]) -> Tuple[List[Dict], List[Dict]]:
|
| 514 |
-
"""بررسی کیفیت ناشناسسازی"""
|
| 515 |
-
correctly_anonymized = []
|
| 516 |
-
missed_entities = []
|
| 517 |
-
|
| 518 |
-
for entity in entities:
|
| 519 |
-
entity_text = entity['text']
|
| 520 |
-
|
| 521 |
-
# بررسی اینکه آیا موجودیت در متن ناشناسسازی شده وجود دارد یا نه
|
| 522 |
-
if entity_text in anonymized_text:
|
| 523 |
-
# موجودیت ناشناسسازی نشده
|
| 524 |
-
missed_entities.append({
|
| 525 |
-
'text': entity_text,
|
| 526 |
-
'category': entity['category'],
|
| 527 |
-
'reason': 'موجود در متن ناشناسسازی شده'
|
| 528 |
-
})
|
| 529 |
-
else:
|
| 530 |
-
# بررسی اینکه آیا جایگزین شده یا حذف شده
|
| 531 |
-
original_words = set(original_text.split())
|
| 532 |
-
anonymized_words = set(anonymized_text.split())
|
| 533 |
-
entity_words = set(entity_text.split())
|
| 534 |
-
|
| 535 |
-
if entity_words.issubset(original_words) and not entity_words.issubset(anonymized_words):
|
| 536 |
-
correctly_anonymized.append(entity)
|
| 537 |
-
else:
|
| 538 |
-
# بررسی دقیقتر با استفاده از similarity
|
| 539 |
-
if self._is_anonymized_with_similarity(original_text, anonymized_text, entity_text):
|
| 540 |
-
correctly_anonymized.append(entity)
|
| 541 |
-
else:
|
| 542 |
-
missed_entities.append({
|
| 543 |
-
'text': entity_text,
|
| 544 |
-
'category': entity['category'],
|
| 545 |
-
'reason': 'تشخیص ناشناسسازی ناموفق'
|
| 546 |
-
})
|
| 547 |
-
|
| 548 |
-
return correctly_anonymized, missed_entities
|
| 549 |
-
|
| 550 |
-
def _is_anonymized_with_similarity(self, original: str, anonymized: str, entity_text: str) -> bool:
|
| 551 |
-
"""بررسی ناشناسسازی با استفاده از شباهت متنی"""
|
| 552 |
-
try:
|
| 553 |
-
# حذف موجودیت از متن اصلی
|
| 554 |
-
original_without_entity = original.replace(entity_text, "[REMOVED]")
|
| 555 |
-
|
| 556 |
-
# محاسبه شباهت
|
| 557 |
-
similarity = SequenceMatcher(None, original_without_entity, anonymized).ratio()
|
| 558 |
-
|
| 559 |
-
# اگر شباهت بالا باشد، احتمالاً ناشناسسازی شده
|
| 560 |
-
return similarity > 0.7
|
| 561 |
-
|
| 562 |
-
except:
|
| 563 |
-
return False
|
| 564 |
-
|
| 565 |
-
def _count_false_positives(self, anonymized_text: str) -> int:
|
| 566 |
-
"""شمارش کلمات اشتباه ناشناسسازی شده"""
|
| 567 |
-
# شمارش کلماتی که به نظر placeholder هستند اما نباید باشند
|
| 568 |
-
false_positive_patterns = [
|
| 569 |
-
r'\b[A-Z_]+_\d+_ANONYMIZED\b',
|
| 570 |
-
r'\[\w+\]',
|
| 571 |
-
r'\*+',
|
| 572 |
-
]
|
| 573 |
-
|
| 574 |
-
false_positives = 0
|
| 575 |
-
for pattern in false_positive_patterns:
|
| 576 |
-
false_positives += len(re.findall(pattern, anonymized_text))
|
| 577 |
-
|
| 578 |
-
return false_positives
|
| 579 |
-
|
| 580 |
-
def _calculate_confidence_score(self, entities: List[Dict], correctly_anonymized: int, total_entities: int) -> float:
|
| 581 |
-
"""محاسبه امتیاز اعتماد"""
|
| 582 |
-
if total_entities == 0:
|
| 583 |
-
return 1.0
|
| 584 |
-
|
| 585 |
-
accuracy = correctly_anonymized / total_entities
|
| 586 |
-
diversity = min(1.0, len(set(e['category'] for e in entities)) / 10)
|
| 587 |
-
|
| 588 |
-
confidence = (accuracy * 0.8 + diversity * 0.2)
|
| 589 |
-
return round(confidence, 3)
|
| 590 |
-
|
| 591 |
-
def _get_memory_usage(self) -> float:
|
| 592 |
-
"""دریافت مصرف حافظه فعلی"""
|
| 593 |
-
if not PSUTIL_AVAILABLE or not self.config.enable_memory_profiling:
|
| 594 |
-
return 0.0
|
| 595 |
-
try:
|
| 596 |
-
process = psutil.Process()
|
| 597 |
-
return process.memory_info().rss / 1024 / 1024 # MB
|
| 598 |
-
except:
|
| 599 |
-
return 0.0
|
| 600 |
-
|
| 601 |
-
# =============================================================================
|
| 602 |
-
# Enhanced Benchmark Interface
|
| 603 |
-
# =============================================================================
|
| 604 |
-
|
| 605 |
-
class EnhancedBenchmarkInterface:
|
| 606 |
-
"""رابط کاربری پیشرفته بنچمارک"""
|
| 607 |
-
|
| 608 |
-
def __init__(self):
|
| 609 |
-
self.current_results = None
|
| 610 |
-
self.current_language = 'fa'
|
| 611 |
-
self.config = BenchmarkConfig()
|
| 612 |
-
|
| 613 |
-
try:
|
| 614 |
-
self.anonymizer = EnhancedComparisonAnonymizer(self.config)
|
| 615 |
-
self.system_ready = True
|
| 616 |
-
logger.info("✅ Enhanced comparison system initialized")
|
| 617 |
-
except Exception as e:
|
| 618 |
-
logger.error(f"❌ System initialization failed: {e}")
|
| 619 |
-
self.system_ready = False
|
| 620 |
-
|
| 621 |
-
def load_local_datasets(self) -> Tuple[List[Tuple[str, str]], List[Tuple[str, str]]]:
|
| 622 |
-
"""بارگذاری دیتاستهای محلی با جفت متون"""
|
| 623 |
-
persian_pairs = []
|
| 624 |
-
english_pairs = []
|
| 625 |
-
|
| 626 |
-
def find_text_columns(df):
|
| 627 |
-
"""پیدا کردن ستونهای متن اصلی و ناشناسسازی شده"""
|
| 628 |
-
# تمیز کردن نام ستونها
|
| 629 |
-
df.columns = df.columns.str.strip()
|
| 630 |
-
|
| 631 |
-
original_col = None
|
| 632 |
-
anonymized_col = None
|
| 633 |
-
|
| 634 |
-
# ستونهای احتمالی برای متن اصلی
|
| 635 |
-
original_candidates = ['original_text', 'original', 'text', 'sentence', 'content', 'input']
|
| 636 |
-
# ستونهای احتمالی برای متن ناشناسسازی شده
|
| 637 |
-
anonymized_candidates = ['anonymized_text', 'anonymized', 'output', 'result', 'processed']
|
| 638 |
-
|
| 639 |
-
logger.info(f"Available columns: {list(df.columns)}")
|
| 640 |
-
|
| 641 |
-
for col in original_candidates:
|
| 642 |
-
if col in df.columns:
|
| 643 |
-
original_col = col
|
| 644 |
-
logger.info(f"Found original text column: {col}")
|
| 645 |
-
break
|
| 646 |
-
|
| 647 |
-
for col in anonymized_candidates:
|
| 648 |
-
if col in df.columns:
|
| 649 |
-
anonymized_col = col
|
| 650 |
-
logger.info(f"Found anonymized text column: {col}")
|
| 651 |
-
break
|
| 652 |
-
|
| 653 |
-
# اگر ستونهای دقیق پیدا نشد، از دو ستون اول استفاده کن
|
| 654 |
-
if not original_col and not anonymized_col and len(df.columns) >= 2:
|
| 655 |
-
original_col = df.columns[0]
|
| 656 |
-
anonymized_col = df.columns[1]
|
| 657 |
-
logger.info(f"Using first two columns: {original_col}, {anonymized_col}")
|
| 658 |
-
|
| 659 |
-
return original_col, anonymized_col
|
| 660 |
-
|
| 661 |
-
try:
|
| 662 |
-
# بارگذاری دیتاست فارسی
|
| 663 |
-
fa_files = ['dataset-fa.csv', 'datasetfa.csv', 'datasetfa.txt', 'dataset_fa.csv']
|
| 664 |
-
for filename in fa_files:
|
| 665 |
-
if os.path.exists(filename):
|
| 666 |
-
try:
|
| 667 |
-
logger.info(f"Attempting to load Persian dataset: {filename}")
|
| 668 |
-
|
| 669 |
-
if filename.endswith('.csv'):
|
| 670 |
-
# تلاش برای خواندن با encoding های مختلف
|
| 671 |
-
df_fa = None
|
| 672 |
-
for encoding in ['utf-8', 'utf-8-sig', 'cp1256', 'iso-8859-1']:
|
| 673 |
-
try:
|
| 674 |
-
df_fa = pd.read_csv(filename, encoding=encoding)
|
| 675 |
-
logger.info(f"Successfully read {filename} with encoding: {encoding}")
|
| 676 |
-
break
|
| 677 |
-
except UnicodeDecodeError:
|
| 678 |
-
continue
|
| 679 |
-
|
| 680 |
-
if df_fa is None:
|
| 681 |
-
logger.error(f"Could not read {filename} with any encoding")
|
| 682 |
-
continue
|
| 683 |
-
|
| 684 |
-
logger.info(f"File shape: {df_fa.shape}")
|
| 685 |
-
logger.info(f"Columns before cleaning: {list(df_fa.columns)}")
|
| 686 |
-
|
| 687 |
-
original_col, anonymized_col = find_text_columns(df_fa)
|
| 688 |
-
|
| 689 |
-
if original_col and anonymized_col:
|
| 690 |
-
valid_pairs = 0
|
| 691 |
-
for _, row in df_fa.iterrows():
|
| 692 |
-
try:
|
| 693 |
-
orig_text = str(row[original_col]).strip()
|
| 694 |
-
anon_text = str(row[anonymized_col]).strip()
|
| 695 |
-
|
| 696 |
-
# بررسی اینکه متنها معتبر هستند
|
| 697 |
-
if (orig_text and anon_text and
|
| 698 |
-
orig_text != 'nan' and anon_text != 'nan' and
|
| 699 |
-
len(orig_text) > 5 and len(anon_text) > 5):
|
| 700 |
-
persian_pairs.append((orig_text, anon_text))
|
| 701 |
-
valid_pairs += 1
|
| 702 |
-
except Exception as e:
|
| 703 |
-
logger.warning(f"Error processing row: {e}")
|
| 704 |
-
continue
|
| 705 |
-
|
| 706 |
-
logger.info(f"✅ Loaded {valid_pairs} Persian pairs from {filename}")
|
| 707 |
-
if valid_pairs > 0:
|
| 708 |
-
break
|
| 709 |
-
else:
|
| 710 |
-
logger.warning(f"❌ Could not find appropriate columns in {filename}")
|
| 711 |
-
logger.warning(f"Available columns: {list(df_fa.columns)}")
|
| 712 |
-
|
| 713 |
-
except Exception as e:
|
| 714 |
-
logger.error(f"Error loading Persian {filename}: {e}")
|
| 715 |
-
continue
|
| 716 |
-
|
| 717 |
-
# بارگذاری دیتاست انگلیسی (همان منطق)
|
| 718 |
-
en_files = ['dataset-en.csv', 'dataseten.csv', 'dataseten.txt', 'dataset_en.csv']
|
| 719 |
-
for filename in en_files:
|
| 720 |
-
if os.path.exists(filename):
|
| 721 |
-
try:
|
| 722 |
-
logger.info(f"Attempting to load English dataset: {filename}")
|
| 723 |
-
|
| 724 |
-
if filename.endswith('.csv'):
|
| 725 |
-
df_en = None
|
| 726 |
-
for encoding in ['utf-8', 'utf-8-sig', 'cp1256', 'iso-8859-1']:
|
| 727 |
-
try:
|
| 728 |
-
df_en = pd.read_csv(filename, encoding=encoding)
|
| 729 |
-
logger.info(f"Successfully read {filename} with encoding: {encoding}")
|
| 730 |
-
break
|
| 731 |
-
except UnicodeDecodeError:
|
| 732 |
-
continue
|
| 733 |
-
|
| 734 |
-
if df_en is None:
|
| 735 |
-
logger.error(f"Could not read {filename} with any encoding")
|
| 736 |
-
continue
|
| 737 |
-
|
| 738 |
-
logger.info(f"File shape: {df_en.shape}")
|
| 739 |
-
logger.info(f"Columns before cleaning: {list(df_en.columns)}")
|
| 740 |
-
|
| 741 |
-
original_col, anonymized_col = find_text_columns(df_en)
|
| 742 |
-
|
| 743 |
-
if original_col and anonymized_col:
|
| 744 |
-
valid_pairs = 0
|
| 745 |
-
for _, row in df_en.iterrows():
|
| 746 |
-
try:
|
| 747 |
-
orig_text = str(row[original_col]).strip()
|
| 748 |
-
anon_text = str(row[anonymized_col]).strip()
|
| 749 |
-
|
| 750 |
-
if (orig_text and anon_text and
|
| 751 |
-
orig_text != 'nan' and anon_text != 'nan' and
|
| 752 |
-
len(orig_text) > 5 and len(anon_text) > 5):
|
| 753 |
-
english_pairs.append((orig_text, anon_text))
|
| 754 |
-
valid_pairs += 1
|
| 755 |
-
except Exception as e:
|
| 756 |
-
logger.warning(f"Error processing row: {e}")
|
| 757 |
-
continue
|
| 758 |
-
|
| 759 |
-
logger.info(f"✅ Loaded {valid_pairs} English pairs from {filename}")
|
| 760 |
-
if valid_pairs > 0:
|
| 761 |
-
break
|
| 762 |
-
else:
|
| 763 |
-
logger.warning(f"❌ Could not find appropriate columns in {filename}")
|
| 764 |
-
logger.warning(f"Available columns: {list(df_en.columns)}")
|
| 765 |
-
|
| 766 |
-
except Exception as e:
|
| 767 |
-
logger.error(f"Error loading English {filename}: {e}")
|
| 768 |
-
continue
|
| 769 |
-
|
| 770 |
-
except Exception as e:
|
| 771 |
-
logger.error(f"❌ Error loading local datasets: {e}")
|
| 772 |
-
|
| 773 |
-
logger.info(f"Final counts - Persian: {len(persian_pairs)}, English: {len(english_pairs)}")
|
| 774 |
-
return persian_pairs, english_pairs
|
| 775 |
-
|
| 776 |
-
def run_enhanced_benchmark(self, sample_size: int, enable_parallel: bool = True,
|
| 777 |
-
enable_clustering: bool = False, progress=gr.Progress()):
|
| 778 |
-
"""اجرای بنچمارک پیشرفته با مقایسه واقعی"""
|
| 779 |
-
|
| 780 |
-
if not self.system_ready:
|
| 781 |
-
return self._get_error_response("System not ready")
|
| 782 |
-
|
| 783 |
-
try:
|
| 784 |
-
progress(0.05, desc="Loading local datasets...")
|
| 785 |
-
|
| 786 |
-
# بارگذاری دیتاستهای محلی
|
| 787 |
-
persian_pairs, english_pairs = self.load_local_datasets()
|
| 788 |
-
|
| 789 |
-
if not persian_pairs and not english_pairs:
|
| 790 |
-
return self._get_error_response("No text pairs loaded from local datasets. Check file format and columns.")
|
| 791 |
-
|
| 792 |
-
# ترکیب جفتها
|
| 793 |
-
all_pairs = persian_pairs + english_pairs
|
| 794 |
-
|
| 795 |
-
# محدود کردن تعداد
|
| 796 |
-
if len(all_pairs) > sample_size:
|
| 797 |
-
all_pairs = all_pairs[:sample_size]
|
| 798 |
-
|
| 799 |
-
# تنظیم پیکربندی
|
| 800 |
-
self.config.sample_size = len(all_pairs)
|
| 801 |
-
self.config.enable_parallel_processing = enable_parallel
|
| 802 |
-
self.config.enable_clustering_analysis = enable_clustering
|
| 803 |
-
|
| 804 |
-
progress(0.1, desc=f"Comparing {len(all_pairs)} text pairs...")
|
| 805 |
-
|
| 806 |
-
# پردازش جفتها
|
| 807 |
-
results = []
|
| 808 |
-
start_time = time.time()
|
| 809 |
-
|
| 810 |
-
if enable_parallel and len(all_pairs) > 10:
|
| 811 |
-
results = self._process_parallel(all_pairs, progress)
|
| 812 |
-
else:
|
| 813 |
-
results = self._process_sequential(all_pairs, progress)
|
| 814 |
-
|
| 815 |
-
total_time = time.time() - start_time
|
| 816 |
-
|
| 817 |
-
# محاسبه آمار کلی
|
| 818 |
-
progress(0.85, desc="Calculating comprehensive metrics...")
|
| 819 |
-
|
| 820 |
-
successful_results = [r for r in results if r.success]
|
| 821 |
-
|
| 822 |
-
if not successful_results:
|
| 823 |
-
return self._get_error_response("No successful results")
|
| 824 |
-
|
| 825 |
-
# محاسبه متریکهای پیشرفته
|
| 826 |
-
summary = self._calculate_comprehensive_metrics(successful_results, total_time)
|
| 827 |
-
|
| 828 |
-
# اجرای تحلیلهای اضافی
|
| 829 |
-
if enable_clustering and SKLEARN_AVAILABLE:
|
| 830 |
-
progress(0.92, desc="Running clustering analysis...")
|
| 831 |
-
clustering_results = self._run_clustering_analysis(successful_results)
|
| 832 |
-
summary['clustering_analysis'] = clustering_results
|
| 833 |
-
|
| 834 |
-
# تست استرس
|
| 835 |
-
progress(0.95, desc="Running stress test...")
|
| 836 |
-
if all_pairs:
|
| 837 |
-
stress_results = self._run_stress_test(all_pairs[0])
|
| 838 |
-
summary['stress_test'] = stress_results
|
| 839 |
-
|
| 840 |
-
# ذخیره نتایج
|
| 841 |
-
self.current_results = {
|
| 842 |
-
'summary': summary,
|
| 843 |
-
'detailed_results': [self._result_to_dict(r) for r in results],
|
| 844 |
-
'timestamp': datetime.now().isoformat(),
|
| 845 |
-
'benchmark_version': 'enhanced_comparison_v3.1',
|
| 846 |
-
'config': {
|
| 847 |
-
'sample_size': sample_size,
|
| 848 |
-
'parallel_processing': enable_parallel,
|
| 849 |
-
'clustering_enabled': enable_clustering,
|
| 850 |
-
}
|
| 851 |
-
}
|
| 852 |
-
|
| 853 |
-
progress(1.0, desc="Enhanced benchmark completed!")
|
| 854 |
-
|
| 855 |
-
# ایجاد گزارش جامع متنی
|
| 856 |
-
detailed_report = self._create_comprehensive_report()
|
| 857 |
-
|
| 858 |
-
success_msg = f"✅ Enhanced benchmark completed! Compared {len(all_pairs)} text pairs with real accuracy metrics"
|
| 859 |
-
|
| 860 |
-
return (
|
| 861 |
-
success_msg,
|
| 862 |
-
detailed_report, # فقط گزارش متنی
|
| 863 |
-
gr.update(visible=True), # results visibility
|
| 864 |
-
gr.update(visible=True), # download button
|
| 865 |
-
)
|
| 866 |
-
|
| 867 |
-
except Exception as e:
|
| 868 |
-
logger.error(f"❌ Benchmark error: {e}")
|
| 869 |
-
return self._get_error_response(f"Benchmark failed: {str(e)}")
|
| 870 |
-
|
| 871 |
-
def _process_sequential(self, pairs: List[Tuple[str, str]], progress) -> List[ComparisonResult]:
|
| 872 |
-
"""پردازش ترتیبی جفتها"""
|
| 873 |
-
results = []
|
| 874 |
-
|
| 875 |
-
for i, (original, anonymized) in enumerate(pairs):
|
| 876 |
-
progress(0.1 + (0.7 * i / len(pairs)),
|
| 877 |
-
desc=f"Comparing pair {i+1}/{len(pairs)}")
|
| 878 |
-
|
| 879 |
-
result = self.anonymizer.compare_texts(original, anonymized)
|
| 880 |
-
results.append(result)
|
| 881 |
-
|
| 882 |
-
if i % 50 == 0: # garbage collection هر 50 جفت
|
| 883 |
-
gc.collect()
|
| 884 |
-
|
| 885 |
-
return results
|
| 886 |
-
|
| 887 |
-
def _process_parallel(self, pairs: List[Tuple[str, str]], progress) -> List[ComparisonResult]:
|
| 888 |
-
"""پردازش موازی جفتها"""
|
| 889 |
-
results = []
|
| 890 |
-
completed = 0
|
| 891 |
-
|
| 892 |
-
max_workers = min(self.config.max_workers, multiprocessing.cpu_count())
|
| 893 |
-
|
| 894 |
-
with ThreadPoolExecutor(max_workers=max_workers) as executor:
|
| 895 |
-
# ارسال وظایف
|
| 896 |
-
future_to_index = {
|
| 897 |
-
executor.submit(self.anonymizer.compare_texts, original, anonymized): i
|
| 898 |
-
for i, (original, anonymized) in enumerate(pairs)
|
| 899 |
-
}
|
| 900 |
-
|
| 901 |
-
# جمعآوری نتایج
|
| 902 |
-
index_to_result = {}
|
| 903 |
-
|
| 904 |
-
for future in as_completed(future_to_index):
|
| 905 |
-
index = future_to_index[future]
|
| 906 |
-
try:
|
| 907 |
-
result = future.result()
|
| 908 |
-
index_to_result[index] = result
|
| 909 |
-
except Exception as e:
|
| 910 |
-
logger.error(f"Error processing pair {index}: {e}")
|
| 911 |
-
original, anonymized = pairs[index] if index < len(pairs) else ("", "")
|
| 912 |
-
index_to_result[index] = ComparisonResult(
|
| 913 |
-
index=index,
|
| 914 |
-
success=False,
|
| 915 |
-
processing_time_ms=0,
|
| 916 |
-
original_text=original[:200],
|
| 917 |
-
anonymized_text=anonymized[:200],
|
| 918 |
-
original_length=len(original),
|
| 919 |
-
anonymized_length=len(anonymized),
|
| 920 |
-
entities_should_anonymize=0,
|
| 921 |
-
entities_correctly_anonymized=0,
|
| 922 |
-
entities_missed=0,
|
| 923 |
-
missed_entities_list=[],
|
| 924 |
-
anonymization_accuracy=0.0,
|
| 925 |
-
precision=0.0,
|
| 926 |
-
recall=0.0,
|
| 927 |
-
f1_score=0.0,
|
| 928 |
-
detected_language='unknown',
|
| 929 |
-
confidence_score=0.0,
|
| 930 |
-
memory_used_mb=0.0,
|
| 931 |
-
entity_categories={},
|
| 932 |
-
error=str(e)
|
| 933 |
-
)
|
| 934 |
-
|
| 935 |
-
completed += 1
|
| 936 |
-
progress(0.1 + (0.7 * completed / len(pairs)),
|
| 937 |
-
desc=f"Completed {completed}/{len(pairs)} pairs")
|
| 938 |
-
|
| 939 |
-
# مرتبسازی نتایج بر اساس ایندکس
|
| 940 |
-
results = [index_to_result[i] for i in sorted(index_to_result.keys())]
|
| 941 |
-
|
| 942 |
-
return results
|
| 943 |
-
|
| 944 |
-
def _calculate_comprehensive_metrics(self, results: List[ComparisonResult], total_time: float) -> Dict:
|
| 945 |
-
"""محاسبه متریکهای جامع بر اساس مقایسه واقعی"""
|
| 946 |
-
|
| 947 |
-
# آمار پایه
|
| 948 |
-
total_pairs = len(results)
|
| 949 |
-
successful_pairs = sum(1 for r in results if r.success)
|
| 950 |
-
success_rate = successful_pairs / total_pairs if total_pairs > 0 else 0
|
| 951 |
-
|
| 952 |
-
processing_times = [r.processing_time_ms for r in results if r.success]
|
| 953 |
-
avg_processing_time = np.mean(processing_times) if processing_times else 0
|
| 954 |
-
|
| 955 |
-
# محاسبه آمار واقعی ناشناسسازی
|
| 956 |
-
total_entities_should_anonymize = sum(r.entities_should_anonymize for r in results if r.success)
|
| 957 |
-
total_correctly_anonymized = sum(r.entities_correctly_anonymized for r in results if r.success)
|
| 958 |
-
total_missed = sum(r.entities_missed for r in results if r.success)
|
| 959 |
-
|
| 960 |
-
# محاسبه متریکهای کلی
|
| 961 |
-
overall_precision = np.mean([r.precision for r in results if r.success and r.precision > 0])
|
| 962 |
-
overall_recall = np.mean([r.recall for r in results if r.success and r.recall > 0])
|
| 963 |
-
overall_f1 = np.mean([r.f1_score for r in results if r.success and r.f1_score > 0])
|
| 964 |
-
overall_anonymization_rate = np.mean([r.anonymization_accuracy for r in results if r.success])
|
| 965 |
-
|
| 966 |
-
# تعدیل نرخ موفقیت بر اساس entities جا افتاده
|
| 967 |
-
if total_entities_should_anonymize > 0:
|
| 968 |
-
entity_miss_penalty = (total_missed / total_entities_should_anonymize) * 100
|
| 969 |
-
adjusted_success_rate = max(0, success_rate * 100 - entity_miss_penalty) / 100
|
| 970 |
-
else:
|
| 971 |
-
adjusted_success_rate = success_rate
|
| 972 |
-
|
| 973 |
-
# محاسبه مقیاسپذیری
|
| 974 |
-
scalability_score = self._calculate_scalability_score(processing_times)
|
| 975 |
-
|
| 976 |
-
# آمار حافظه
|
| 977 |
-
memory_usage = [r.memory_used_mb for r in results if r.success]
|
| 978 |
-
memory_stats = {
|
| 979 |
-
'avg_memory_per_pair': np.mean(memory_usage) if memory_usage else 0,
|
| 980 |
-
'peak_memory': max(memory_usage) if memory_usage else 0,
|
| 981 |
-
'total_memory': sum(memory_usage) if memory_usage else 0,
|
| 982 |
-
}
|
| 983 |
-
|
| 984 |
-
# آمار زبان
|
| 985 |
-
languages = [r.detected_language for r in results if r.success and r.detected_language]
|
| 986 |
-
language_distribution = Counter(languages)
|
| 987 |
-
|
| 988 |
-
# تحلیل دستههای جا افتاده
|
| 989 |
-
missed_categories = Counter()
|
| 990 |
-
for result in results:
|
| 991 |
-
if result.success and result.missed_entities_list:
|
| 992 |
-
for missed_entity in result.missed_entities_list:
|
| 993 |
-
missed_categories[missed_entity['category']] += 1
|
| 994 |
-
|
| 995 |
-
# امتیاز کلی عملکرد بر اساس مقایسه واقعی
|
| 996 |
-
performance_score = self._calculate_realistic_performance_score(
|
| 997 |
-
adjusted_success_rate, avg_processing_time, overall_anonymization_rate, scalability_score
|
| 998 |
-
)
|
| 999 |
-
|
| 1000 |
-
return {
|
| 1001 |
-
'total_pairs': total_pairs,
|
| 1002 |
-
'successful_pairs': successful_pairs,
|
| 1003 |
-
'success_rate': success_rate,
|
| 1004 |
-
'adjusted_success_rate': adjusted_success_rate, # تعدیل شده بر اساس entities جا افتاده
|
| 1005 |
-
'avg_processing_time_ms': avg_processing_time,
|
| 1006 |
-
'total_entities_should_anonymize': total_entities_should_anonymize,
|
| 1007 |
-
'total_correctly_anonymized': total_correctly_anonymized,
|
| 1008 |
-
'total_missed_entities': total_missed,
|
| 1009 |
-
'overall_anonymization_rate': overall_anonymization_rate,
|
| 1010 |
-
'pairs_per_minute': (successful_pairs / max(0.01, total_time / 60)),
|
| 1011 |
-
'total_benchmark_time': total_time,
|
| 1012 |
-
'scalability_score': scalability_score,
|
| 1013 |
-
'performance_score': performance_score,
|
| 1014 |
-
'memory_stats': memory_stats,
|
| 1015 |
-
'language_distribution': dict(language_distribution),
|
| 1016 |
-
'missed_categories_analysis': dict(missed_categories.most_common(10)),
|
| 1017 |
-
'quality_metrics': {
|
| 1018 |
-
'precision': round(overall_precision if not np.isnan(overall_precision) else 0, 1),
|
| 1019 |
-
'recall': round(overall_recall if not np.isnan(overall_recall) else 0, 1),
|
| 1020 |
-
'f1_score': round(overall_f1 if not np.isnan(overall_f1) else 0, 1),
|
| 1021 |
-
'accuracy': round(overall_anonymization_rate * 100, 1)
|
| 1022 |
-
},
|
| 1023 |
-
'processing_time_percentiles': {
|
| 1024 |
-
'50th': np.percentile(processing_times, 50) if processing_times else 0,
|
| 1025 |
-
'95th': np.percentile(processing_times, 95) if processing_times else 0,
|
| 1026 |
-
'99th': np.percentile(processing_times, 99) if processing_times else 0,
|
| 1027 |
-
},
|
| 1028 |
-
'entity_miss_penalty_percent': (total_missed / max(1, total_entities_should_anonymize)) * 100
|
| 1029 |
-
}
|
| 1030 |
-
|
| 1031 |
-
def _calculate_realistic_performance_score(self, adjusted_success_rate: float, avg_time: float,
|
| 1032 |
-
anonymization_rate: float, scalability: float) -> float:
|
| 1033 |
-
"""محاسبه امتیاز کلی عملکرد بر اساس مقایسه واقعی"""
|
| 1034 |
-
|
| 1035 |
-
weights = {
|
| 1036 |
-
'adjusted_success': 0.35, # وزن بیشتر برای نرخ موفقیت تعدیل شده
|
| 1037 |
-
'anonymization_quality': 0.30, # کیفیت ناشناسسازی واقعی
|
| 1038 |
-
'speed': 0.20,
|
| 1039 |
-
'scalability': 0.15
|
| 1040 |
-
}
|
| 1041 |
-
|
| 1042 |
-
success_score = adjusted_success_rate * 100
|
| 1043 |
-
quality_score = anonymization_rate * 100
|
| 1044 |
-
speed_score = min(100, max(0, 100 - (avg_time / 10))) # کمتر بهتر
|
| 1045 |
-
scalability_score = scalability
|
| 1046 |
-
|
| 1047 |
-
total_score = (
|
| 1048 |
-
weights['adjusted_success'] * success_score +
|
| 1049 |
-
weights['anonymization_quality'] * quality_score +
|
| 1050 |
-
weights['speed'] * speed_score +
|
| 1051 |
-
weights['scalability'] * scalability_score
|
| 1052 |
-
)
|
| 1053 |
-
|
| 1054 |
-
return round(total_score, 1)
|
| 1055 |
-
|
| 1056 |
-
def _calculate_scalability_score(self, processing_times: List[float]) -> float:
|
| 1057 |
-
"""محاسبه امتیاز مقیاسپذیری"""
|
| 1058 |
-
if not processing_times:
|
| 1059 |
-
return 100.0
|
| 1060 |
-
|
| 1061 |
-
std_dev = np.std(processing_times)
|
| 1062 |
-
mean_time = np.mean(processing_times)
|
| 1063 |
-
cv = std_dev / mean_time if mean_time > 0 else 1
|
| 1064 |
-
|
| 1065 |
-
scalability = max(0, 100 - (cv * 100))
|
| 1066 |
-
return round(scalability, 1)
|
| 1067 |
-
|
| 1068 |
-
def _run_clustering_analysis(self, results: List[ComparisonResult]) -> Dict:
|
| 1069 |
-
"""اجرای تحلیل کلاسترینگ"""
|
| 1070 |
-
if not SKLEARN_AVAILABLE:
|
| 1071 |
-
return {}
|
| 1072 |
-
|
| 1073 |
-
try:
|
| 1074 |
-
features = []
|
| 1075 |
-
for result in results:
|
| 1076 |
-
feature_vector = [
|
| 1077 |
-
result.processing_time_ms,
|
| 1078 |
-
result.original_length,
|
| 1079 |
-
result.entities_should_anonymize,
|
| 1080 |
-
result.anonymization_accuracy,
|
| 1081 |
-
result.precision,
|
| 1082 |
-
result.recall
|
| 1083 |
-
]
|
| 1084 |
-
features.append(feature_vector)
|
| 1085 |
-
|
| 1086 |
-
if len(features) < 3:
|
| 1087 |
-
return {}
|
| 1088 |
-
|
| 1089 |
-
kmeans = KMeans(n_clusters=min(3, len(features)), random_state=42)
|
| 1090 |
-
clusters = kmeans.fit_predict(features)
|
| 1091 |
-
|
| 1092 |
-
cluster_analysis = {}
|
| 1093 |
-
for i in range(max(clusters) + 1):
|
| 1094 |
-
cluster_results = [results[j] for j, c in enumerate(clusters) if c == i]
|
| 1095 |
-
cluster_analysis[f'cluster_{i}'] = {
|
| 1096 |
-
'count': len(cluster_results),
|
| 1097 |
-
'avg_processing_time': np.mean([r.processing_time_ms for r in cluster_results]),
|
| 1098 |
-
'avg_entities': np.mean([r.entities_should_anonymize for r in cluster_results]),
|
| 1099 |
-
'avg_accuracy': np.mean([r.anonymization_accuracy for r in cluster_results])
|
| 1100 |
-
}
|
| 1101 |
-
|
| 1102 |
-
return cluster_analysis
|
| 1103 |
-
|
| 1104 |
-
except Exception as e:
|
| 1105 |
-
logger.error(f"Clustering analysis failed: {e}")
|
| 1106 |
-
return {}
|
| 1107 |
-
|
| 1108 |
-
def _run_stress_test(self, sample_pair: Tuple[str, str]) -> Dict:
|
| 1109 |
-
"""اجرای تست استرس"""
|
| 1110 |
-
try:
|
| 1111 |
-
iterations = self.config.stress_test_iterations
|
| 1112 |
-
response_times = []
|
| 1113 |
-
successful = 0
|
| 1114 |
-
failed = 0
|
| 1115 |
-
|
| 1116 |
-
original, anonymized = sample_pair
|
| 1117 |
-
|
| 1118 |
-
for _ in range(iterations):
|
| 1119 |
-
start_time = time.time()
|
| 1120 |
-
try:
|
| 1121 |
-
result = self.anonymizer.compare_texts(original, anonymized)
|
| 1122 |
-
if result.success:
|
| 1123 |
-
successful += 1
|
| 1124 |
-
else:
|
| 1125 |
-
failed += 1
|
| 1126 |
-
except:
|
| 1127 |
-
failed += 1
|
| 1128 |
-
|
| 1129 |
-
response_time = (time.time() - start_time) * 1000
|
| 1130 |
-
response_times.append(response_time)
|
| 1131 |
-
|
| 1132 |
-
return {
|
| 1133 |
-
'iterations': iterations,
|
| 1134 |
-
'successful': successful,
|
| 1135 |
-
'failed': failed,
|
| 1136 |
-
'avg_response_time': np.mean(response_times),
|
| 1137 |
-
'max_response_time': max(response_times),
|
| 1138 |
-
'min_response_time': min(response_times),
|
| 1139 |
-
'throughput_per_sec': 1000 / np.mean(response_times) if response_times else 0
|
| 1140 |
-
}
|
| 1141 |
-
|
| 1142 |
-
except Exception as e:
|
| 1143 |
-
logger.error(f"Stress test failed: {e}")
|
| 1144 |
-
return {}
|
| 1145 |
-
|
| 1146 |
-
def _result_to_dict(self, result: ComparisonResult) -> Dict:
|
| 1147 |
-
"""تبدیل ComparisonResult به دیکشنری"""
|
| 1148 |
-
return {
|
| 1149 |
-
'index': result.index,
|
| 1150 |
-
'success': result.success,
|
| 1151 |
-
'processing_time_ms': result.processing_time_ms,
|
| 1152 |
-
'original_length': result.original_length,
|
| 1153 |
-
'anonymized_length': result.anonymized_length,
|
| 1154 |
-
'entities_should_anonymize': result.entities_should_anonymize,
|
| 1155 |
-
'entities_correctly_anonymized': result.entities_correctly_anonymized,
|
| 1156 |
-
'entities_missed': result.entities_missed,
|
| 1157 |
-
'missed_entities_list': result.missed_entities_list,
|
| 1158 |
-
'anonymization_accuracy': result.anonymization_accuracy,
|
| 1159 |
-
'precision': result.precision,
|
| 1160 |
-
'recall': result.recall,
|
| 1161 |
-
'f1_score': result.f1_score,
|
| 1162 |
-
'detected_language': result.detected_language,
|
| 1163 |
-
'confidence_score': result.confidence_score,
|
| 1164 |
-
'memory_used_mb': result.memory_used_mb,
|
| 1165 |
-
'entity_categories': result.entity_categories,
|
| 1166 |
-
'error': result.error,
|
| 1167 |
-
}
|
| 1168 |
-
|
| 1169 |
-
def _create_comprehensive_report(self) -> str:
|
| 1170 |
-
"""ایجاد گزارش جامع و تفصیلی"""
|
| 1171 |
-
if not self.current_results:
|
| 1172 |
-
return "No results available."
|
| 1173 |
-
|
| 1174 |
-
summary = self.current_results['summary']
|
| 1175 |
-
config = self.current_results['config']
|
| 1176 |
-
|
| 1177 |
-
# تحلیل نتایج تفصیلی
|
| 1178 |
-
detailed_results = self.current_results['detailed_results']
|
| 1179 |
-
successful_results = [r for r in detailed_results if r.get('success', False)]
|
| 1180 |
-
|
| 1181 |
-
# آمار تفصیلی
|
| 1182 |
-
processing_times = [r['processing_time_ms'] for r in successful_results]
|
| 1183 |
-
accuracy_scores = [r['anonymization_accuracy'] for r in successful_results]
|
| 1184 |
-
precision_scores = [r['precision'] for r in successful_results if r['precision'] > 0]
|
| 1185 |
-
recall_scores = [r['recall'] for r in successful_results if r['recall'] > 0]
|
| 1186 |
-
|
| 1187 |
-
# تحلیل دستهبندیها
|
| 1188 |
-
all_categories = defaultdict(int)
|
| 1189 |
-
missed_categories = defaultdict(int)
|
| 1190 |
-
|
| 1191 |
-
for result in successful_results:
|
| 1192 |
-
for category, count in result.get('entity_categories', {}).items():
|
| 1193 |
-
all_categories[category] += count
|
| 1194 |
-
|
| 1195 |
-
for missed_entity in result.get('missed_entities_list', []):
|
| 1196 |
-
missed_categories[missed_entity['category']] += 1
|
| 1197 |
-
|
| 1198 |
-
report = f"""
|
| 1199 |
-
# 📊 گزارش جامع بنچمارک مقایسه واقعی - نسخه 3.1
|
| 1200 |
-
|
| 1201 |
-
## ⭐ خلاصه کلیدی
|
| 1202 |
-
**امتیاز عملکرد کلی:** {summary['performance_score']:.1f}/100
|
| 1203 |
-
**نرخ موفقیت تعدیل شده:** {summary['adjusted_success_rate']*100:.1f}%
|
| 1204 |
-
**دقت ناشناسسازی واقعی:** {summary['overall_anonymization_rate']*100:.1f}%
|
| 1205 |
-
**جریمه entities جا افتاده:** {summary['entity_miss_penalty_percent']:.1f}%
|
| 1206 |
-
|
| 1207 |
-
## 📈 آمار کلی عملیات
|
| 1208 |
-
- **کل جفت متون بررسی شده**: {summary['total_pairs']:,}
|
| 1209 |
-
- **جفتهای موفق**: {summary['successful_pairs']:,}
|
| 1210 |
-
- **نرخ موفقیت اولیه**: {summary['success_rate']*100:.1f}%
|
| 1211 |
-
- **نرخ موفقیت تعدیل شده**: {summary['adjusted_success_rate']*100:.1f}% (پس از کسر جریمه entities جا افتاده)
|
| 1212 |
-
|
| 1213 |
-
## 🎯 تحلیل عمیق ناشناسسازی واقعی
|
| 1214 |
-
|
| 1215 |
-
### نتایج اصلی:
|
| 1216 |
-
- **کل entities که باید ناشناسسازی میشدند**: {summary['total_entities_should_anonymize']:,}
|
| 1217 |
-
- **entities صحیح ناشناسسازی شده**: {summary['total_correctly_anonymized']:,}
|
| 1218 |
-
- **entities جا افتاده**: {summary['total_missed_entities']:,}
|
| 1219 |
-
- **نرخ دقت واقعی**: {summary['overall_anonymization_rate']*100:.1f}%
|
| 1220 |
-
|
| 1221 |
-
### متریکهای کیفیت پیشرفته:
|
| 1222 |
-
- **دقت (Precision)**: {summary['quality_metrics']['precision']:.1f}%
|
| 1223 |
-
- **بازخوانی (Recall)**: {summary['quality_metrics']['recall']:.1f}%
|
| 1224 |
-
- **امتیاز F1**: {summary['quality_metrics']['f1_score']:.1f}%
|
| 1225 |
-
- **صحت کلی (Accuracy)**: {summary['quality_metrics']['accuracy']:.1f}%
|
| 1226 |
-
|
| 1227 |
-
## ⚡ آنالیز عملکرد سیستم
|
| 1228 |
-
|
| 1229 |
-
### سرعت و توان عملیاتی:
|
| 1230 |
-
- **متوسط زمان پردازش هر جفت**: {summary['avg_processing_time_ms']:.1f} میلیثانیه
|
| 1231 |
-
- **توان عملیاتی**: {summary['pairs_per_minute']:.0f} جفت/دقیقه
|
| 1232 |
-
- **کل زمان بنچمارک**: {summary['total_benchmark_time']:.1f} ثانیه
|
| 1233 |
-
- **امتیاز مقیاسپذیری**: {summary['scalability_score']:.1f}/100
|
| 1234 |
-
|
| 1235 |
-
### توزیع زمان پردازش:
|
| 1236 |
-
"""
|
| 1237 |
-
|
| 1238 |
-
if processing_times:
|
| 1239 |
-
report += f"""- **میانه زمان پردازش**: {np.median(processing_times):.1f} ms
|
| 1240 |
-
- **حداقل زمان**: {min(processing_times):.1f} ms
|
| 1241 |
-
- **حداکثر زمان**: {max(processing_times):.1f} ms
|
| 1242 |
-
- **انحراف معیار**: {np.std(processing_times):.1f} ms
|
| 1243 |
-
"""
|
| 1244 |
-
|
| 1245 |
-
percentiles = summary.get('processing_time_percentiles', {})
|
| 1246 |
-
if percentiles:
|
| 1247 |
-
report += f"""- **50درصدی زمان**: {percentiles.get('50th', 0):.1f} ms
|
| 1248 |
-
- **95درصدی زمان**: {percentiles.get('95th', 0):.1f} ms
|
| 1249 |
-
- **99درصدی زمان**: {percentiles.get('99th', 0):.1f} ms
|
| 1250 |
-
"""
|
| 1251 |
-
|
| 1252 |
-
report += f"""
|
| 1253 |
-
## 💾 تحلیل مصرف حافظه
|
| 1254 |
-
- **متوسط حافظه هر جفت**: {summary['memory_stats']['avg_memory_per_pair']:.2f} MB
|
| 1255 |
-
- **حداکثر مصرف حافظه**: {summary['memory_stats']['peak_memory']:.2f} MB
|
| 1256 |
-
- **کل حافظه استفاده شده**: {summary['memory_stats']['total_memory']:.2f} MB
|
| 1257 |
-
- **امتیاز کارایی حافظه**: {max(0, 100 - summary['memory_stats']['avg_memory_per_pair']):.1f}/100
|
| 1258 |
-
|
| 1259 |
-
## 🌍 تحلیل زبانها
|
| 1260 |
-
"""
|
| 1261 |
-
|
| 1262 |
-
lang_dist = summary['language_distribution']
|
| 1263 |
-
total_samples = sum(lang_dist.values())
|
| 1264 |
-
for lang, count in lang_dist.items():
|
| 1265 |
-
lang_name = {'fa': 'فارسی', 'en': 'انگلیسی', 'mixed': 'ترکیبی', 'unknown': 'نامشخص'}.get(lang, lang)
|
| 1266 |
-
percentage = (count/total_samples)*100 if total_samples > 0 else 0
|
| 1267 |
-
report += f"- **{lang_name}**: {count} جفت ({percentage:.1f}%)\n"
|
| 1268 |
-
|
| 1269 |
-
report += f"""
|
| 1270 |
-
## 🔍 تحلیل عمیق دستههای entities
|
| 1271 |
-
|
| 1272 |
-
### توزیع کل دستهها:
|
| 1273 |
-
"""
|
| 1274 |
-
|
| 1275 |
-
for category, count in sorted(all_categories.items(), key=lambda x: x[1], reverse=True):
|
| 1276 |
-
percentage = (count/sum(all_categories.values()))*100 if sum(all_categories.values()) > 0 else 0
|
| 1277 |
-
report += f"- **{category}**: {count} مورد ({percentage:.1f}%)\n"
|
| 1278 |
-
|
| 1279 |
-
report += f"""
|
| 1280 |
-
### تحلیل دستههای جا افتاده:
|
| 1281 |
-
"""
|
| 1282 |
-
|
| 1283 |
-
if missed_categories:
|
| 1284 |
-
for category, missed_count in sorted(missed_categories.items(), key=lambda x: x[1], reverse=True):
|
| 1285 |
-
total_in_category = all_categories.get(category, 0)
|
| 1286 |
-
miss_rate = (missed_count/total_in_category)*100 if total_in_category > 0 else 0
|
| 1287 |
-
report += f"- **{category}**: {missed_count} جا افتاده از {total_in_category} ({miss_rate:.1f}% نرخ جا افتادگی)\n"
|
| 1288 |
-
else:
|
| 1289 |
-
report += "- هیچ entity جا نیافتاده! (عملکرد فوقالعاده)\n"
|
| 1290 |
-
|
| 1291 |
-
report += f"""
|
| 1292 |
-
## 📊 تحلیل آماری پیشرفته
|
| 1293 |
-
|
| 1294 |
-
### توزیع دقت ناشناسسازی:
|
| 1295 |
-
"""
|
| 1296 |
-
|
| 1297 |
-
if accuracy_scores:
|
| 1298 |
-
report += f"""- **میانگین دقت**: {np.mean(accuracy_scores)*100:.1f}%
|
| 1299 |
-
- **میانه دقت**: {np.median(accuracy_scores)*100:.1f}%
|
| 1300 |
-
- **حداقل دقت**: {min(accuracy_scores)*100:.1f}%
|
| 1301 |
-
- **حداکثر دقت**: {max(accuracy_scores)*100:.1f}%
|
| 1302 |
-
- **انحراف معیار**: {np.std(accuracy_scores)*100:.1f}%
|
| 1303 |
-
- **نمونههای با دقت 100%**: {sum(1 for s in accuracy_scores if s >= 1.0)} از {len(accuracy_scores)}
|
| 1304 |
-
- **نمونههای با دقت کمتر از 50%**: {sum(1 for s in accuracy_scores if s < 0.5)} از {len(accuracy_scores)}
|
| 1305 |
-
"""
|
| 1306 |
-
|
| 1307 |
-
# تحلیل کارایی بر اساس طول متن
|
| 1308 |
-
if successful_results:
|
| 1309 |
-
long_texts = [r for r in successful_results if r['original_length'] > 200]
|
| 1310 |
-
short_texts = [r for r in successful_results if r['original_length'] <= 200]
|
| 1311 |
-
|
| 1312 |
-
if long_texts and short_texts:
|
| 1313 |
-
long_avg_accuracy = np.mean([r['anonymization_accuracy'] for r in long_texts])
|
| 1314 |
-
short_avg_accuracy = np.mean([r['anonymization_accuracy'] for r in short_texts])
|
| 1315 |
-
long_avg_time = np.mean([r['processing_time_ms'] for r in long_texts])
|
| 1316 |
-
short_avg_time = np.mean([r['processing_time_ms'] for r in short_texts])
|
| 1317 |
-
|
| 1318 |
-
report += f"""
|
| 1319 |
-
### تحلیل بر اساس طول متن:
|
| 1320 |
-
- **متون کوتاه (≤200 کاراکتر)**: {len(short_texts)} نمونه
|
| 1321 |
-
- میانگین دقت: {short_avg_accuracy*100:.1f}%
|
| 1322 |
-
- میانگین زمان: {short_avg_time:.1f} ms
|
| 1323 |
-
- **متون طولانی (>200 کاراکتر)**: {len(long_texts)} نمونه
|
| 1324 |
-
- میانگین دقت: {long_avg_accuracy*100:.1f}%
|
| 1325 |
-
- میانگین زمان: {long_avg_time:.1f} ms
|
| 1326 |
-
"""
|
| 1327 |
-
|
| 1328 |
-
stress = summary.get('stress_test', {})
|
| 1329 |
-
if stress and 'iterations' in stress:
|
| 1330 |
-
report += f"""
|
| 1331 |
-
## 🔥 نتایج تست استرس
|
| 1332 |
-
- **کل تکرارها**: {stress['iterations']}
|
| 1333 |
-
- **موفق**: {stress['successful']} ({stress['successful']/stress['iterations']*100:.1f}%)
|
| 1334 |
-
- **ناموفق**: {stress['failed']} ({stress['failed']/stress['iterations']*100:.1f}%)
|
| 1335 |
-
- **متوسط زمان پاسخ**: {stress['avg_response_time']:.1f} ms
|
| 1336 |
-
- **حداکثر زمان پاسخ**: {stress['max_response_time']:.1f} ms
|
| 1337 |
-
- **حداقل زمان پاسخ**: {stress['min_response_time']:.1f} ms
|
| 1338 |
-
- **توان عملیاتی**: {stress['throughput_per_sec']:.1f} عملیات/ثانیه
|
| 1339 |
-
"""
|
| 1340 |
-
|
| 1341 |
-
# ارزیابی و پیشنهادات
|
| 1342 |
-
performance = summary['performance_score']
|
| 1343 |
-
miss_penalty = summary['entity_miss_penalty_percent']
|
| 1344 |
-
|
| 1345 |
-
if performance >= 85 and miss_penalty < 10:
|
| 1346 |
-
report += """
|
| 1347 |
-
## ✅ سیستم شما عملکرد عالی دارد!
|
| 1348 |
-
|
| 1349 |
-
### نقاط قوت:
|
| 1350 |
-
- ناشناسسازی با دقت بالا انجام میشود
|
| 1351 |
-
- entities جا افتاده کم است
|
| 1352 |
-
- سرعت پردازش مطلوب
|
| 1353 |
-
- مقیاسپذیری خوب
|
| 1354 |
-
- استفاده بهینه از حافظه
|
| 1355 |
-
|
| 1356 |
-
### توصیهها:
|
| 1357 |
-
- آماده برای استفاده در محیط تولید
|
| 1358 |
-
- ادامه مانیتورینگ منظم
|
| 1359 |
-
- حفظ کیفیت فعلی در بهروزرسانیها
|
| 1360 |
-
- تست دورهای با دادههای جدید
|
| 1361 |
-
"""
|
| 1362 |
-
elif performance >= 70 or miss_penalty < 20:
|
| 1363 |
-
report += """
|
| 1364 |
-
## ⚠️ سیستم نیاز به بهبودهایی دارد:
|
| 1365 |
-
|
| 1366 |
-
### مشکلات شناسایی شده:
|
| 1367 |
-
- نرخ entities جا افتاده قابل توجه است
|
| 1368 |
-
- دقت ناشناسسازی نیاز به بهبود دارد
|
| 1369 |
-
- عملکرد در برخی دستهها ضعیف است
|
| 1370 |
-
|
| 1371 |
-
### پیشنهادات بهبود:
|
| 1372 |
-
- بهینهسازی الگوریتمهای تشخیص موجودیت
|
| 1373 |
-
- کاهش entities جا افتاده با تنظیم threshold ها
|
| 1374 |
-
- بهبود دقت pattern matching
|
| 1375 |
-
- افزایش پوشش دستههای مختلف
|
| 1376 |
-
- تنظیم دقیقتر پارامترهای سیستم
|
| 1377 |
-
- آموزش مجدد مدلها با دادههای بیشتر
|
| 1378 |
-
|
| 1379 |
-
### اولویتهای فوری:
|
| 1380 |
-
1. کاهش نرخ جا افتادگی در دستههای حساس
|
| 1381 |
-
2. بهبود دقت تشخیص entities
|
| 1382 |
-
3. تست با حجم بیشتر داده
|
| 1383 |
-
"""
|
| 1384 |
-
else:
|
| 1385 |
-
report += """
|
| 1386 |
-
## 🔧 سیستم نیاز به بازنگری اساسی دارد:
|
| 1387 |
-
|
| 1388 |
-
### مشکلات جدی شناسایی شده:
|
| 1389 |
-
- نرخ بالای entities جا افتاده
|
| 1390 |
-
- دقت ناشناسسازی پایین
|
| 1391 |
-
- عملکرد غیرقابل اعتماد
|
| 1392 |
-
|
| 1393 |
-
### اقدامات ضروری:
|
| 1394 |
-
- بازطراحی کامل الگوریتمهای اصلی
|
| 1395 |
-
- افزایش چشمگیر patterns تشخیص
|
| 1396 |
-
- پیادهسازی مدلهای NER بهتر
|
| 1397 |
-
- بهبود preprocessing متون
|
| 1398 |
-
- training مجدد با دادههای بیشتر
|
| 1399 |
-
- کاهش قابل توجه نرخ false negative ها
|
| 1400 |
-
- تست گسترده قبل از استقرار
|
| 1401 |
-
|
| 1402 |
-
### پیشنهاد:
|
| 1403 |
-
سیستم در حال حاضر آماده استقرار در محیط تولید نیست.
|
| 1404 |
-
نیاز به توسعه اساسی و تستهای گستردهتر دارد.
|
| 1405 |
-
"""
|
| 1406 |
-
|
| 1407 |
-
report += f"""
|
| 1408 |
-
|
| 1409 |
-
## 📋 تنظیمات بنچمارک
|
| 1410 |
-
- **اندازه نمونه**: {config['sample_size']}
|
| 1411 |
-
- **پردازش موازی**: {'فعال' if config['parallel_processing'] else 'غیرفعال'}
|
| 1412 |
-
- **تحلیل کلاسترینگ**: {'فعال' if config['clustering_enabled'] else 'غیرفعال'}
|
| 1413 |
-
|
| 1414 |
-
## 📌 نتیجهگیری
|
| 1415 |
-
این گزارش بر اساس مقایسه واقعی {summary['total_pairs']} جفت متن تولید شده است.
|
| 1416 |
-
نتایج نشاندهنده عملکرد واقعی سیستم ناشناسسازی با در نظر گیری
|
| 1417 |
-
entities جا افتاده و دقت فعلی الگوریتمهای تشخیص میباشد.
|
| 1418 |
-
|
| 1419 |
-
**زمان تولید گزارش**: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}
|
| 1420 |
-
**نسخه بنچمارک**: Enhanced Real Comparison v3.1
|
| 1421 |
-
|
| 1422 |
-
---
|
| 1423 |
-
*این گزارش برای ارزیابی دقیق و بهبود سیستم ناشناسسازی طراحی شده است*
|
| 1424 |
-
"""
|
| 1425 |
-
|
| 1426 |
-
return report
|
| 1427 |
-
|
| 1428 |
-
def _get_error_response(self, error_msg: str):
|
| 1429 |
-
"""پاسخ استاندارد برای خطاها"""
|
| 1430 |
-
return (
|
| 1431 |
-
f"❌ {error_msg}",
|
| 1432 |
-
"خطا در اجرای benchmark رخ داده است.",
|
| 1433 |
-
gr.update(visible=False),
|
| 1434 |
-
gr.update(visible=False)
|
| 1435 |
-
)
|
| 1436 |
-
|
| 1437 |
-
def download_results(self):
|
| 1438 |
-
"""دانلود نتایج"""
|
| 1439 |
-
if not self.current_results:
|
| 1440 |
-
return None
|
| 1441 |
-
|
| 1442 |
-
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
| 1443 |
-
filename = f"real_comparison_benchmark_{timestamp}.json"
|
| 1444 |
-
|
| 1445 |
-
with open(filename, 'w', encoding='utf-8') as f:
|
| 1446 |
-
json.dump(self.current_results, f, ensure_ascii=False, indent=2, default=str)
|
| 1447 |
-
|
| 1448 |
-
return filename
|
| 1449 |
-
|
| 1450 |
-
# =============================================================================
|
| 1451 |
-
# Main Interface Creation
|
| 1452 |
-
# =============================================================================
|
| 1453 |
-
|
| 1454 |
-
def create_enhanced_interface():
|
| 1455 |
-
"""ایجاد رابط کاربری پیشرفته"""
|
| 1456 |
-
|
| 1457 |
-
benchmark_interface = EnhancedBenchmarkInterface()
|
| 1458 |
-
|
| 1459 |
-
custom_css = """
|
| 1460 |
-
.gradio-container {
|
| 1461 |
-
font-family: 'Segoe UI', Tahoma, Arial, sans-serif !important;
|
| 1462 |
-
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%) !important;
|
| 1463 |
-
min-height: 100vh !important;
|
| 1464 |
-
}
|
| 1465 |
-
|
| 1466 |
-
.gradio-button {
|
| 1467 |
-
border-radius: 25px !important;
|
| 1468 |
-
font-weight: bold !important;
|
| 1469 |
-
transition: all 0.3s ease !important;
|
| 1470 |
-
margin: 5px 0 !important;
|
| 1471 |
-
min-height: 50px !important;
|
| 1472 |
-
}
|
| 1473 |
-
|
| 1474 |
-
.gradio-button:hover {
|
| 1475 |
-
transform: translateY(-2px) !important;
|
| 1476 |
-
box-shadow: 0 6px 20px rgba(0,0,0,0.2) !important;
|
| 1477 |
-
}
|
| 1478 |
-
|
| 1479 |
-
h1, h2, h3 {
|
| 1480 |
-
text-shadow: 2px 2px 4px rgba(0,0,0,0.3) !important;
|
| 1481 |
-
margin: 10px 0 !important;
|
| 1482 |
-
line-height: 1.2 !important;
|
| 1483 |
-
}
|
| 1484 |
-
"""
|
| 1485 |
-
|
| 1486 |
-
with gr.Blocks(title="🚀 Enhanced Real Comparison Benchmark", theme=gr.themes.Soft(), css=custom_css) as app:
|
| 1487 |
-
|
| 1488 |
-
# Header
|
| 1489 |
-
gr.HTML("""
|
| 1490 |
-
<div style="text-align: center; padding: 20px;">
|
| 1491 |
-
<h1>🚀 Enhanced Real Comparison Anonymization Benchmark v3.1</h1>
|
| 1492 |
-
<p style="font-size: 1.2rem;">Real Performance Analysis with Original vs Anonymized Text Comparison</p>
|
| 1493 |
-
<p>✨ Features: Real Entity Counting, Miss Rate Calculation, Adjusted Success Metrics</p>
|
| 1494 |
-
</div>
|
| 1495 |
-
""")
|
| 1496 |
-
|
| 1497 |
-
with gr.Row():
|
| 1498 |
-
with gr.Column(scale=1):
|
| 1499 |
-
# Dataset Status
|
| 1500 |
-
gr.HTML("""
|
| 1501 |
-
<div style="background-color: #e8f4f8; padding: 15px; border-radius: 10px; margin-bottom: 15px;">
|
| 1502 |
-
<h3>📁 Dataset Requirements</h3>
|
| 1503 |
-
<p>System looks for files with columns:</p>
|
| 1504 |
-
<ul>
|
| 1505 |
-
<li><strong>original_text</strong> - Original sentences</li>
|
| 1506 |
-
<li><strong>anonymized_text</strong> - Anonymized versions</li>
|
| 1507 |
-
</ul>
|
| 1508 |
-
<p><strong>Supported files:</strong></p>
|
| 1509 |
-
<p>datasetfa.csv, dataset-fa.csv, dataseten.csv, dataset-en.csv</p>
|
| 1510 |
-
</div>
|
| 1511 |
-
""")
|
| 1512 |
-
|
| 1513 |
-
sample_size = gr.Slider(
|
| 1514 |
-
minimum=10,
|
| 1515 |
-
maximum=500,
|
| 1516 |
-
value=100,
|
| 1517 |
-
step=5,
|
| 1518 |
-
label="📊 Sample Size (Text Pairs to Compare)"
|
| 1519 |
-
)
|
| 1520 |
-
|
| 1521 |
-
with gr.Row():
|
| 1522 |
-
enable_parallel = gr.Checkbox(
|
| 1523 |
-
label="⚡ Enable Parallel Processing",
|
| 1524 |
-
value=True
|
| 1525 |
-
)
|
| 1526 |
-
|
| 1527 |
-
enable_clustering = gr.Checkbox(
|
| 1528 |
-
label="🎯 Enable Clustering Analysis",
|
| 1529 |
-
value=False
|
| 1530 |
-
)
|
| 1531 |
-
|
| 1532 |
-
run_btn = gr.Button(
|
| 1533 |
-
"🚀 Run Real Comparison Benchmark",
|
| 1534 |
-
variant="primary",
|
| 1535 |
-
size="lg"
|
| 1536 |
-
)
|
| 1537 |
-
|
| 1538 |
-
download_btn = gr.Button(
|
| 1539 |
-
"📥 Download Complete Results",
|
| 1540 |
-
variant="secondary",
|
| 1541 |
-
visible=False
|
| 1542 |
-
)
|
| 1543 |
-
|
| 1544 |
-
status_output = gr.Textbox(
|
| 1545 |
-
label="📋 Status",
|
| 1546 |
-
interactive=False,
|
| 1547 |
-
lines=4
|
| 1548 |
-
)
|
| 1549 |
-
|
| 1550 |
-
with gr.Column(scale=2):
|
| 1551 |
-
# Results - Only Text Report
|
| 1552 |
-
detailed_report = gr.Markdown(
|
| 1553 |
-
"No results yet. Please run the real comparison benchmark first.",
|
| 1554 |
-
visible=False,
|
| 1555 |
-
label="📊 Comprehensive Analysis Report"
|
| 1556 |
-
)
|
| 1557 |
-
|
| 1558 |
-
# System Status with file check
|
| 1559 |
-
system_status = "✅ Enhanced comparison system ready" if benchmark_interface.system_ready else "⚠️ Running in limited mode"
|
| 1560 |
-
|
| 1561 |
-
# Check for datasets
|
| 1562 |
-
fa_files = ['datasetfa.csv', 'dataset-fa.csv']
|
| 1563 |
-
en_files = ['dataseten.csv', 'dataset-en.csv']
|
| 1564 |
-
|
| 1565 |
-
fa_status = "❌ Not found"
|
| 1566 |
-
en_status = "❌ Not found"
|
| 1567 |
-
|
| 1568 |
-
for f in fa_files:
|
| 1569 |
-
if os.path.exists(f):
|
| 1570 |
-
fa_status = f"✅ Found: {f}"
|
| 1571 |
-
break
|
| 1572 |
-
|
| 1573 |
-
for f in en_files:
|
| 1574 |
-
if os.path.exists(f):
|
| 1575 |
-
en_status = f"✅ Found: {f}"
|
| 1576 |
-
break
|
| 1577 |
-
|
| 1578 |
-
gr.HTML(f"""
|
| 1579 |
-
<div style="text-align: center; margin-top: 20px; padding: 15px; background-color: #e8f4f8; border-radius: 10px;">
|
| 1580 |
-
<p><strong>System Status:</strong> {system_status}</p>
|
| 1581 |
-
<p><strong>Persian Dataset:</strong> {fa_status}</p>
|
| 1582 |
-
<p><strong>English Dataset:</strong> {en_status}</p>
|
| 1583 |
-
<p><strong>Available Features:</strong>
|
| 1584 |
-
{'✅ Parallel Processing' if multiprocessing.cpu_count() > 1 else '❌ Sequential Only'} |
|
| 1585 |
-
{'✅ Advanced Metrics' if SKLEARN_AVAILABLE else '❌ Basic Metrics'} |
|
| 1586 |
-
{'✅ Memory Profiling' if PSUTIL_AVAILABLE else '❌ No Memory Tracking'} |
|
| 1587 |
-
{'✅ NER Models' if SPACY_AVAILABLE else '❌ Pattern-Only'}
|
| 1588 |
-
</p>
|
| 1589 |
-
</div>
|
| 1590 |
-
""")
|
| 1591 |
-
|
| 1592 |
-
# Usage Guide
|
| 1593 |
-
with gr.Accordion("📖 Real Comparison Guide", open=False):
|
| 1594 |
-
gr.HTML("""
|
| 1595 |
-
<div style="padding: 20px;">
|
| 1596 |
-
<h3>🚀 How It Works</h3>
|
| 1597 |
-
<ol>
|
| 1598 |
-
<li><strong>Dataset Preparation:</strong> CSV files with original_text and anonymized_text columns</li>
|
| 1599 |
-
<li><strong>Real Entity Detection:</strong> System finds entities that should be anonymized in original text</li>
|
| 1600 |
-
<li><strong>Comparison Analysis:</strong> Checks which entities were actually anonymized</li>
|
| 1601 |
-
<li><strong>Miss Rate Calculation:</strong> Calculates percentage of missed entities</li>
|
| 1602 |
-
<li><strong>Adjusted Metrics:</strong> Success rate adjusted based on missed entities</li>
|
| 1603 |
-
</ol>
|
| 1604 |
-
|
| 1605 |
-
<h3>📊 Key Metrics</h3>
|
| 1606 |
-
<ul>
|
| 1607 |
-
<li><strong>Real Anonymization Accuracy:</strong> Correctly anonymized / Should be anonymized</li>
|
| 1608 |
-
<li><strong>Adjusted Success Rate:</strong> Initial success - entity miss penalty</li>
|
| 1609 |
-
<li><strong>Entity Miss Penalty:</strong> (Missed entities / Total entities) × 100</li>
|
| 1610 |
-
<li><strong>Precision & Recall:</strong> Based on actual entity detection and anonymization</li>
|
| 1611 |
-
</ul>
|
| 1612 |
-
|
| 1613 |
-
<h3>🎯 Expected File Format</h3>
|
| 1614 |
-
<pre>
|
| 1615 |
-
original_text,anonymized_text
|
| 1616 |
-
"John Smith works at Apple Inc","PERSON_001 works at COMPANY_001"
|
| 1617 |
-
"Call me at 555-1234","Call me at PHONE_001"
|
| 1618 |
-
</pre>
|
| 1619 |
-
</div>
|
| 1620 |
-
""")
|
| 1621 |
-
|
| 1622 |
-
# Event Handlers
|
| 1623 |
-
run_btn.click(
|
| 1624 |
-
fn=benchmark_interface.run_enhanced_benchmark,
|
| 1625 |
-
inputs=[sample_size, enable_parallel, enable_clustering],
|
| 1626 |
-
outputs=[
|
| 1627 |
-
status_output,
|
| 1628 |
-
detailed_report,
|
| 1629 |
-
download_btn,
|
| 1630 |
-
detailed_report # visibility toggle
|
| 1631 |
-
],
|
| 1632 |
-
show_progress=True
|
| 1633 |
-
)
|
| 1634 |
-
|
| 1635 |
-
download_btn.click(
|
| 1636 |
-
fn=benchmark_interface.download_results,
|
| 1637 |
-
outputs=gr.File()
|
| 1638 |
-
)
|
| 1639 |
-
|
| 1640 |
-
return app
|
| 1641 |
-
|
| 1642 |
-
# =============================================================================
|
| 1643 |
-
# Main Function
|
| 1644 |
-
# =============================================================================
|
| 1645 |
-
|
| 1646 |
-
def main():
|
| 1647 |
-
"""تابع اصلی"""
|
| 1648 |
-
|
| 1649 |
-
print("🚀 Starting Enhanced Real Comparison Benchmark System v3.1...")
|
| 1650 |
-
print("=" * 80)
|
| 1651 |
-
|
| 1652 |
-
# Check for datasets
|
| 1653 |
-
datasets = [
|
| 1654 |
-
('dataset-fa.csv', 'Persian'),
|
| 1655 |
-
('dataset-fa.csv', 'Persian (Alt)'),
|
| 1656 |
-
('dataset-en.csv', 'English'),
|
| 1657 |
-
('dataset-en.csv', 'English (Alt)')
|
| 1658 |
-
]
|
| 1659 |
-
|
| 1660 |
-
for filename, desc in datasets:
|
| 1661 |
-
if os.path.exists(filename):
|
| 1662 |
-
print(f"✅ {desc} dataset found: {filename}")
|
| 1663 |
-
try:
|
| 1664 |
-
df = pd.read_csv(filename, encoding='utf-8')
|
| 1665 |
-
print(f" - Rows: {len(df)}")
|
| 1666 |
-
print(f" - Columns: {list(df.columns)}")
|
| 1667 |
-
if 'original_text' in df.columns and 'anonymized_text' in df.columns:
|
| 1668 |
-
print(f" - ✅ Required columns present")
|
| 1669 |
-
else:
|
| 1670 |
-
print(f" - ❌ Missing required columns: original_text, anonymized_text")
|
| 1671 |
-
except Exception as e:
|
| 1672 |
-
print(f" - ❌ Error reading file: {e}")
|
| 1673 |
-
else:
|
| 1674 |
-
print(f"❌ {desc} dataset not found: {filename}")
|
| 1675 |
-
|
| 1676 |
-
# System capabilities
|
| 1677 |
-
features = []
|
| 1678 |
-
if SKLEARN_AVAILABLE:
|
| 1679 |
-
features.append("Advanced ML Metrics")
|
| 1680 |
-
if PSUTIL_AVAILABLE:
|
| 1681 |
-
features.append("Memory Profiling")
|
| 1682 |
-
if SPACY_AVAILABLE:
|
| 1683 |
-
features.append("NER Models")
|
| 1684 |
-
|
| 1685 |
-
features.extend(["Real Entity Comparison", "Miss Rate Analysis", "Adjusted Metrics"])
|
| 1686 |
-
|
| 1687 |
-
print(f"✨ Available features: {', '.join(features)}")
|
| 1688 |
-
print(f"🖥️ CPU Cores: {multiprocessing.cpu_count()}")
|
| 1689 |
-
print(f"🧠 Memory: {psutil.virtual_memory().total // (1024**3) if PSUTIL_AVAILABLE else 'Unknown'} GB")
|
| 1690 |
-
|
| 1691 |
-
# Create and launch interface
|
| 1692 |
-
demo = create_enhanced_interface()
|
| 1693 |
-
|
| 1694 |
-
# Launch with enhanced configuration
|
| 1695 |
-
demo.launch(
|
| 1696 |
-
server_name="0.0.0.0",
|
| 1697 |
-
server_port=7860,
|
| 1698 |
-
share=False,
|
| 1699 |
-
inbrowser=True,
|
| 1700 |
-
show_error=True,
|
| 1701 |
-
favicon_path=None,
|
| 1702 |
-
ssl_verify=False,
|
| 1703 |
-
max_file_size="50mb"
|
| 1704 |
-
)
|
| 1705 |
-
|
| 1706 |
-
if __name__ == "__main__":
|
| 1707 |
-
main()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|