Spaces:
Running
Running
File size: 12,020 Bytes
e7a189a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 |
import pandas as pd
import json
import re
import hashlib
from typing import List, Dict, Tuple
from collections import Counter
import unicodedata
from datetime import datetime
class DataQualityManager:
"""Data Quality Management and Standardization"""
def __init__(self):
self.quality_report = {}
self.cleaned_data = []
def clean_text(self, text: str) -> str:
"""Clean and normalize Thai text"""
if not text or not isinstance(text, str):
return ""
# Remove HTML tags
text = re.sub(r'<[^>]+>', '', text)
# Remove excessive whitespace
text = re.sub(r'\s+', ' ', text)
# Normalize Thai characters
text = unicodedata.normalize('NFC', text)
# Clean Thai specific issues
text = re.sub(r'ๆ+', 'ๆ', text) # Multiple repetition marks
text = re.sub(r'[฿๏๎๚๛]', '', text) # Remove special Thai symbols
# Remove URLs
text = re.sub(r'http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\\(\\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+', '', text)
# Remove email addresses
text = re.sub(r'\S+@\S+', '', text)
return text.strip()
def detect_duplicates(self, data: List[Dict]) -> Tuple[List[int], Dict]:
"""Detect duplicate records"""
seen_hashes = {}
duplicates = []
for i, record in enumerate(data):
# Create hash from input content
content = str(record.get('prompt', '')) + str(record.get('input', ''))
content_hash = hashlib.md5(content.encode()).hexdigest()
if content_hash in seen_hashes:
duplicates.append(i)
else:
seen_hashes[content_hash] = i
return duplicates, {"total_duplicates": len(duplicates), "unique_records": len(seen_hashes)}
def validate_completeness(self, data: List[Dict]) -> Dict:
"""Check data completeness"""
required_fields = ['id', 'prompt', 'generated_text']
incomplete_records = []
for i, record in enumerate(data):
missing_fields = [field for field in required_fields if not record.get(field)]
if missing_fields:
incomplete_records.append({
'record_id': i,
'missing_fields': missing_fields
})
return {
"incomplete_records": len(incomplete_records),
"details": incomplete_records[:10] # Show first 10
}
def analyze_quality_metrics(self, data: List[Dict]) -> Dict:
"""Analyze various quality metrics"""
if not data:
return {}
# Text length statistics
prompt_lengths = [len(str(record.get('prompt', ''))) for record in data]
output_lengths = [len(str(record.get('generated_text', ''))) for record in data]
# Language detection (simplified for Thai)
thai_pattern = re.compile(r'[ก-๏]')
thai_records = sum(1 for record in data if thai_pattern.search(str(record.get('generated_text', ''))))
# Model distribution
model_usage = Counter([record.get('model_used', 'unknown') for record in data])
return {
"total_records": len(data),
"avg_prompt_length": sum(prompt_lengths) / len(prompt_lengths) if prompt_lengths else 0,
"avg_output_length": sum(output_lengths) / len(output_lengths) if output_lengths else 0,
"thai_content_ratio": thai_records / len(data) if data else 0,
"model_distribution": dict(model_usage),
"length_stats": {
"min_prompt": min(prompt_lengths) if prompt_lengths else 0,
"max_prompt": max(prompt_lengths) if prompt_lengths else 0,
"min_output": min(output_lengths) if output_lengths else 0,
"max_output": max(output_lengths) if output_lengths else 0
}
}
def standardize_format(self, data: List[Dict], task_type: str) -> Tuple[List[Dict], Dict]:
"""Standardize dataset format according to international standards"""
standardized_data = []
for i, record in enumerate(data):
# Create standardized record
std_record = {
"id": f"{task_type}_{i+1:06d}",
"task_type": task_type,
"input": self.clean_text(str(record.get('prompt', ''))),
"output": self.clean_text(str(record.get('generated_text', ''))),
"metadata": {
"model_used": record.get('model_used', 'unknown'),
"generation_time": record.get('generation_time'),
"language": "th",
"domain": self._detect_domain(record),
"quality_score": self._calculate_quality_score(record)
}
}
# Add original data if available
if record.get('original_data'):
std_record["metadata"]["source_data"] = record['original_data']
standardized_data.append(std_record)
# Create dataset metadata
dataset_metadata = {
"dataset_name": f"thai_{task_type}_dataset",
"created_at": datetime.now().isoformat(),
"version": "1.0.0",
"language": "th",
"task_type": task_type,
"total_samples": len(standardized_data),
"license": "CC-BY-4.0",
"description": f"High-quality Thai {task_type} dataset generated using multiple language models"
}
return standardized_data, dataset_metadata
def _detect_domain(self, record: Dict) -> str:
"""Detect domain/topic of the record"""
text = str(record.get('prompt', '')) + str(record.get('generated_text', ''))
text_lower = text.lower()
# Simple domain detection
if any(word in text_lower for word in ['สุขภาพ', 'โรค', 'ยา', 'แพทย์']):
return "health"
elif any(word in text_lower for word in ['การศึกษา', 'โรงเรียน', 'นักเรียน']):
return "education"
elif any(word in text_lower for word in ['เทคโนโลยี', 'คอมพิวเตอร์', 'โปรแกรม']):
return "technology"
elif any(word in text_lower for word in ['การเงิน', 'ธนาคาร', 'เงิน']):
return "finance"
else:
return "general"
def _calculate_quality_score(self, record: Dict) -> float:
"""Calculate quality score for a record (0-1)"""
score = 1.0
prompt = str(record.get('prompt', ''))
output = str(record.get('generated_text', ''))
# Penalize very short outputs
if len(output) < 10:
score -= 0.3
# Penalize repetitive content
if len(set(output.split())) / len(output.split()) < 0.7 if output.split() else True:
score -= 0.2
# Penalize incomplete responses
if output.endswith('...') or len(output) < len(prompt) * 0.5:
score -= 0.2
# Bonus for Thai content
thai_pattern = re.compile(r'[ก-๏]')
if thai_pattern.search(output):
score += 0.1
return max(0.0, min(1.0, score))
def create_data_splits(self, data: List[Dict], train_ratio: float = 0.8,
val_ratio: float = 0.1, test_ratio: float = 0.1) -> Dict:
"""Create train/validation/test splits"""
import random
# Shuffle data
shuffled_data = data.copy()
random.shuffle(shuffled_data)
total = len(shuffled_data)
train_end = int(total * train_ratio)
val_end = train_end + int(total * val_ratio)
return {
"train": shuffled_data[:train_end],
"validation": shuffled_data[train_end:val_end],
"test": shuffled_data[val_end:]
}
def generate_dataset_card(self, metadata: Dict, quality_metrics: Dict) -> str:
"""Generate dataset card (README) in markdown format"""
card_template = f"""# Thai {metadata['task_type'].title()} Dataset
## Dataset Description
This is a high-quality Thai {metadata['task_type']} dataset created using multiple state-of-the-art language models.
## Dataset Information
- **Language**: Thai (th)
- **Task Type**: {metadata['task_type']}
- **Total Samples**: {metadata['total_samples']:,}
- **Created**: {metadata['created_at']}
- **Version**: {metadata['version']}
- **License**: {metadata['license']}
## Quality Metrics
- **Average Prompt Length**: {quality_metrics.get('avg_prompt_length', 0):.1f} characters
- **Average Output Length**: {quality_metrics.get('avg_output_length', 0):.1f} characters
- **Thai Content Ratio**: {quality_metrics.get('thai_content_ratio', 0):.2%}
## Model Distribution
{self._format_model_distribution(quality_metrics.get('model_distribution', {}))}
## Data Fields
- `id`: Unique identifier for each sample
- `task_type`: Type of NLP task
- `input`: Input prompt or question
- `output`: Generated response or answer
- `metadata`: Additional information including model used, quality score, etc.
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("path/to/dataset")
```
## License
This dataset is released under {metadata['license']} license.
## Citation
If you use this dataset in your research, please cite:
```bibtex
@dataset{{thai_{metadata['task_type']}_dataset,
title={{Thai {metadata['task_type'].title()} Dataset}},
author={{Thai Dataset Generator}},
year={{{datetime.now().year}}},
version={{{metadata['version']}}},
url={{https://github.com/your-repo/thai-dataset}}
}}
```
"""
return card_template
def _format_model_distribution(self, model_dist: Dict) -> str:
"""Format model distribution for markdown"""
if not model_dist:
return "No model distribution data available."
lines = []
for model, count in model_dist.items():
lines.append(f"- **{model}**: {count:,} samples")
return "\n".join(lines)
def export_to_huggingface_format(data_splits: Dict, metadata: Dict, output_dir: str):
"""Export dataset in Hugging Face compatible format"""
import os
import json
# Create output directory
os.makedirs(output_dir, exist_ok=True)
# Save data splits
for split_name, split_data in data_splits.items():
with open(os.path.join(output_dir, f"{split_name}.jsonl"), 'w', encoding='utf-8') as f:
for record in split_data:
f.write(json.dumps(record, ensure_ascii=False) + '\n')
# Save dataset info
dataset_info = {
"dataset_name": metadata["dataset_name"],
"config_name": "default",
"version": {"version_str": metadata["version"]},
"description": metadata["description"],
"homepage": "",
"license": metadata["license"],
"features": {
"id": {"dtype": "string"},
"task_type": {"dtype": "string"},
"input": {"dtype": "string"},
"output": {"dtype": "string"},
"metadata": {"dtype": "string"}
},
"splits": {
split_name: {"name": split_name, "num_examples": len(split_data)}
for split_name, split_data in data_splits.items()
}
}
with open(os.path.join(output_dir, "dataset_info.json"), 'w', encoding='utf-8') as f:
json.dump(dataset_info, f, ensure_ascii=False, indent=2)
print(f"Dataset exported to {output_dir}")
|