File size: 12,990 Bytes
6c482f9 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 |
"""
Domain Dataset Module for Cross-Domain Uncertainty Quantification
This module provides functionality for loading and managing datasets from different domains
for evaluating uncertainty quantification methods across domains.
"""
import os
import json
import pandas as pd
import numpy as np
from typing import List, Dict, Any, Union, Optional, Tuple
from datasets import load_dataset
class DomainDataset:
"""Base class for domain-specific datasets."""
def __init__(self, name: str, domain: str):
"""
Initialize the domain dataset.
Args:
name: Name of the dataset
domain: Domain category (e.g., 'medical', 'legal', 'general')
"""
self.name = name
self.domain = domain
self.data = None
def load(self) -> None:
"""Load the dataset."""
raise NotImplementedError("Subclasses must implement this method")
def get_samples(self, n: Optional[int] = None) -> List[Dict[str, Any]]:
"""
Get samples from the dataset.
Args:
n: Number of samples to return (None for all)
Returns:
List of samples with prompts and expected outputs
"""
raise NotImplementedError("Subclasses must implement this method")
def get_prompt_template(self) -> str:
"""
Get the prompt template for this domain.
Returns:
Prompt template string
"""
raise NotImplementedError("Subclasses must implement this method")
class MedicalQADataset(DomainDataset):
"""Dataset for medical question answering."""
def __init__(self, data_path: Optional[str] = None):
"""
Initialize the medical QA dataset.
Args:
data_path: Path to the dataset file (None to use default)
"""
super().__init__("medical_qa", "medical")
self.data_path = data_path
def load(self) -> None:
"""Load the medical QA dataset."""
if self.data_path and os.path.exists(self.data_path):
# Load from local file if available
if self.data_path.endswith('.csv'):
self.data = pd.read_csv(self.data_path)
elif self.data_path.endswith('.json'):
with open(self.data_path, 'r') as f:
self.data = json.load(f)
else:
raise ValueError(f"Unsupported file format: {self.data_path}")
else:
# Use a sample of the MedMCQA dataset from Hugging Face
try:
dataset = load_dataset("medmcqa", split="train[:100]")
self.data = dataset.to_pandas()
except Exception as e:
# Fallback to synthetic data if dataset loading fails
print(f"Failed to load MedMCQA dataset: {e}")
self.data = self._create_synthetic_data()
def _create_synthetic_data(self) -> pd.DataFrame:
"""Create synthetic medical QA data for testing."""
questions = [
"What are the common symptoms of myocardial infarction?",
"How does insulin regulate blood glucose levels?",
"What is the mechanism of action for ACE inhibitors?",
"What are the diagnostic criteria for rheumatoid arthritis?",
"How does the SARS-CoV-2 virus enter human cells?",
"What are the main side effects of chemotherapy?",
"How does the blood-brain barrier function?",
"What is the pathophysiology of type 2 diabetes?",
"How do vaccines create immunity?",
"What are the stages of chronic kidney disease?"
]
# Create a dataframe with questions only (answers would be generated by LLMs)
return pd.DataFrame({
'question': questions,
'domain': ['medical'] * len(questions)
})
def get_samples(self, n: Optional[int] = None) -> List[Dict[str, Any]]:
"""
Get samples from the medical QA dataset.
Args:
n: Number of samples to return (None for all)
Returns:
List of samples with prompts
"""
if self.data is None:
self.load()
if 'question' in self.data.columns:
questions = self.data['question'].tolist()
elif 'question_text' in self.data.columns:
questions = self.data['question_text'].tolist()
else:
raise ValueError("Dataset does not contain question column")
if n is not None:
questions = questions[:n]
# Create samples with prompts
samples = []
for question in questions:
prompt = self.get_prompt_template().format(question=question)
samples.append({
'domain': 'medical',
'question': question,
'prompt': prompt
})
return samples
def get_prompt_template(self) -> str:
"""
Get the prompt template for medical domain.
Returns:
Prompt template string
"""
return "You are a medical expert. Please answer the following medical question accurately and concisely:\n\n{question}"
class LegalQADataset(DomainDataset):
"""Dataset for legal question answering."""
def __init__(self, data_path: Optional[str] = None):
"""
Initialize the legal QA dataset.
Args:
data_path: Path to the dataset file (None to use default)
"""
super().__init__("legal_qa", "legal")
self.data_path = data_path
def load(self) -> None:
"""Load the legal QA dataset."""
if self.data_path and os.path.exists(self.data_path):
# Load from local file if available
if self.data_path.endswith('.csv'):
self.data = pd.read_csv(self.data_path)
elif self.data_path.endswith('.json'):
with open(self.data_path, 'r') as f:
self.data = json.load(f)
else:
raise ValueError(f"Unsupported file format: {self.data_path}")
else:
# Use synthetic data for legal domain
self.data = self._create_synthetic_data()
def _create_synthetic_data(self) -> pd.DataFrame:
"""Create synthetic legal QA data for testing."""
questions = [
"What constitutes a breach of contract?",
"How is intellectual property protected under international law?",
"What are the elements of negligence in tort law?",
"How does the doctrine of stare decisis function in common law systems?",
"What rights are protected under the Fourth Amendment?",
"What is the difference between a patent and a copyright?",
"How does arbitration differ from litigation?",
"What constitutes insider trading under securities law?",
"What are the legal requirements for a valid will?",
"How does diplomatic immunity work under international law?"
]
# Create a dataframe with questions only
return pd.DataFrame({
'question': questions,
'domain': ['legal'] * len(questions)
})
def get_samples(self, n: Optional[int] = None) -> List[Dict[str, Any]]:
"""
Get samples from the legal QA dataset.
Args:
n: Number of samples to return (None for all)
Returns:
List of samples with prompts
"""
if self.data is None:
self.load()
questions = self.data['question'].tolist()
if n is not None:
questions = questions[:n]
# Create samples with prompts
samples = []
for question in questions:
prompt = self.get_prompt_template().format(question=question)
samples.append({
'domain': 'legal',
'question': question,
'prompt': prompt
})
return samples
def get_prompt_template(self) -> str:
"""
Get the prompt template for legal domain.
Returns:
Prompt template string
"""
return "You are a legal expert. Please answer the following legal question accurately and concisely:\n\n{question}"
class GeneralKnowledgeDataset(DomainDataset):
"""Dataset for general knowledge question answering."""
def __init__(self, data_path: Optional[str] = None):
"""
Initialize the general knowledge dataset.
Args:
data_path: Path to the dataset file (None to use default)
"""
super().__init__("general_knowledge", "general")
self.data_path = data_path
def load(self) -> None:
"""Load the general knowledge dataset."""
if self.data_path and os.path.exists(self.data_path):
# Load from local file if available
if self.data_path.endswith('.csv'):
self.data = pd.read_csv(self.data_path)
elif self.data_path.endswith('.json'):
with open(self.data_path, 'r') as f:
self.data = json.load(f)
else:
raise ValueError(f"Unsupported file format: {self.data_path}")
else:
# Use a sample of the TriviaQA dataset from Hugging Face
try:
dataset = load_dataset("trivia_qa", "unfiltered", split="train[:100]")
self.data = dataset.to_pandas()
except Exception as e:
# Fallback to synthetic data if dataset loading fails
print(f"Failed to load TriviaQA dataset: {e}")
self.data = self._create_synthetic_data()
def _create_synthetic_data(self) -> pd.DataFrame:
"""Create synthetic general knowledge data for testing."""
questions = [
"What is the capital of France?",
"Who wrote the novel '1984'?",
"What is the chemical symbol for gold?",
"Which planet is known as the Red Planet?",
"Who painted the Mona Lisa?",
"What is the largest ocean on Earth?",
"What year did World War II end?",
"What is the tallest mountain in the world?",
"Who was the first person to step on the moon?",
"What is the speed of light in a vacuum?"
]
# Create a dataframe with questions only
return pd.DataFrame({
'question': questions,
'domain': ['general'] * len(questions)
})
def get_samples(self, n: Optional[int] = None) -> List[Dict[str, Any]]:
"""
Get samples from the general knowledge dataset.
Args:
n: Number of samples to return (None for all)
Returns:
List of samples with prompts
"""
if self.data is None:
self.load()
if 'question' in self.data.columns:
questions = self.data['question'].tolist()
elif 'question_text' in self.data.columns:
questions = self.data['question_text'].tolist()
else:
raise ValueError("Dataset does not contain question column")
if n is not None:
questions = questions[:n]
# Create samples with prompts
samples = []
for question in questions:
prompt = self.get_prompt_template().format(question=question)
samples.append({
'domain': 'general',
'question': question,
'prompt': prompt
})
return samples
def get_prompt_template(self) -> str:
"""
Get the prompt template for general knowledge domain.
Returns:
Prompt template string
"""
return "Please answer the following general knowledge question accurately and concisely:\n\n{question}"
# Factory function to create domain datasets
def create_domain_dataset(domain: str, data_path: Optional[str] = None) -> DomainDataset:
"""
Create a domain dataset based on the specified domain.
Args:
domain: Domain category ('medical', 'legal', 'general')
data_path: Path to the dataset file (None to use default)
Returns:
Domain dataset instance
"""
if domain == "medical":
return MedicalQADataset(data_path)
elif domain == "legal":
return LegalQADataset(data_path)
elif domain == "general":
return GeneralKnowledgeDataset(data_path)
else:
raise ValueError(f"Unsupported domain: {domain}")
|