File size: 21,255 Bytes
dcc24f8 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 |
"""
Email Classification Module.
This module provides production-grade email classification using both
rule-based and LLM-based approaches. It categorizes emails into predefined
categories with confidence scoring and transaction detection.
Categories:
- finance: Bank transactions, investments, payments
- shopping: E-commerce, orders, deliveries
- work: Job-related, meetings, projects
- newsletter: Digests, articles, subscriptions
- promotional: Marketing, offers, discounts
- social: Social networks, personal messages
- other: Uncategorized emails
Example:
>>> from src.data.classifier import EmailClassifier
>>> classifier = EmailClassifier()
>>> result = classifier.classify(
... subject="Transaction Alert",
... sender="HDFC Bank",
... body="Rs.500 debited from your account"
... )
>>> print(result.category)
'finance'
>>> print(result.is_transaction)
True
Author: Ranjit Behera
License: MIT
"""
from __future__ import annotations
import json
import logging
import re
from dataclasses import dataclass, asdict
from enum import Enum
from typing import (
Any,
ClassVar,
Dict,
List,
Optional,
Tuple,
Union,
)
# Configure module logger
logger = logging.getLogger(__name__)
class EmailCategory(Enum):
"""
Enumeration of email categories.
Each category represents a distinct type of email with specific
characteristics and handling requirements.
Attributes:
FINANCE: Bank and financial transaction emails.
SHOPPING: E-commerce and order-related emails.
WORK: Professional and job-related emails.
NEWSLETTER: News, articles, and subscription content.
PROMOTIONAL: Marketing and advertising emails.
SOCIAL: Social network and personal communication.
OTHER: Emails that don't fit other categories.
"""
FINANCE = "finance"
SHOPPING = "shopping"
WORK = "work"
NEWSLETTER = "newsletter"
PROMOTIONAL = "promotional"
SOCIAL = "social"
OTHER = "other"
@classmethod
def from_string(cls, value: str) -> EmailCategory:
"""
Convert string to EmailCategory enum.
Args:
value: Category name as string.
Returns:
EmailCategory: Corresponding enum value.
Raises:
ValueError: If value doesn't match any category.
"""
try:
return cls(value.lower())
except ValueError:
logger.warning(f"Unknown category '{value}', defaulting to OTHER")
return cls.OTHER
@dataclass
class ClassificationResult:
"""
Result of email classification.
Contains the predicted category, confidence level, reasoning,
and whether the email is a financial transaction.
Attributes:
category: Predicted email category.
confidence: Confidence level ('high', 'medium', 'low').
reason: Human-readable explanation for classification.
is_transaction: True if email is a financial transaction.
scores: Optional dict of category scores for debugging.
Example:
>>> result = ClassificationResult(
... category="finance",
... confidence="high",
... reason="Contains debit keywords and amount",
... is_transaction=True
... )
>>> result.to_dict()
{'category': 'finance', 'confidence': 'high', ...}
"""
category: str
confidence: str
reason: str
is_transaction: bool = False
scores: Optional[Dict[str, float]] = None
# Validation
VALID_CONFIDENCE_LEVELS: ClassVar[set] = {"high", "medium", "low"}
def __post_init__(self) -> None:
"""Validate classification result."""
if self.confidence not in self.VALID_CONFIDENCE_LEVELS:
logger.warning(f"Invalid confidence '{self.confidence}', setting to 'low'")
self.confidence = "low"
def to_dict(self) -> Dict[str, Any]:
"""
Convert result to dictionary.
Returns:
Dict[str, Any]: Classification result as dictionary.
"""
result = asdict(self)
if self.scores is None:
del result['scores']
return result
def to_json(self) -> str:
"""
Convert result to JSON string.
Returns:
str: JSON representation.
"""
return json.dumps(self.to_dict(), indent=2)
class EmailClassifier:
"""
Production-grade email classifier with rule-based and LLM support.
This classifier uses a sophisticated pattern matching system to
categorize emails and detect financial transactions. It can optionally
use an LLM for more nuanced classification.
Features:
- Multi-pattern rule-based classification
- Sender-based categorization
- Transaction detection
- Confidence scoring
- Optional LLM integration
Attributes:
use_llm: Whether to use LLM for classification.
model: LLM model instance (if use_llm=True).
tokenizer: LLM tokenizer (if use_llm=True).
Example:
>>> classifier = EmailClassifier()
>>> result = classifier.classify(
... subject="Your order has shipped",
... sender="Amazon.in",
... body="Your order #123 is on the way"
... )
>>> print(result.category)
'shopping'
Note:
For production use, the rule-based classifier is recommended
due to its speed and consistency. LLM mode requires additional
dependencies and model loading time.
"""
# Transaction detection keywords
TRANSACTION_KEYWORDS: ClassVar[set] = {
'debited', 'credited', 'transaction', 'transfer', 'payment',
'withdrawn', 'deposited', 'paid', 'received', 'upi', 'neft', 'imps',
'rtgs', 'mandate', 'autopay', 'emi', 'refund', 'cashback'
}
# Category patterns with senders and keywords
CATEGORY_PATTERNS: ClassVar[Dict[EmailCategory, Dict[str, List[str]]]] = {
EmailCategory.FINANCE: {
'senders': [
'hdfc', 'icici', 'sbi', 'axis', 'kotak', 'pnb', 'bob',
'canara', 'union bank', 'idbi', 'yes bank', 'indusind',
'bank', 'banking', 'credit card', 'mutual fund', 'zerodha',
'groww', 'upstox', 'cred', 'slice', 'paytm', 'phonepe', 'gpay'
],
'keywords': [
'transaction', 'statement', 'balance', 'debited', 'credited',
'payment', 'transfer', 'upi', 'neft', 'imps', 'account',
'investment', 'dividend', 'interest', 'emi', 'loan', 'credit',
'debit', 'mandate', 'autopay', 'sip', 'mutual fund'
],
},
EmailCategory.SHOPPING: {
'senders': [
'amazon', 'flipkart', 'myntra', 'ajio', 'nykaa', 'meesho',
'bigbasket', 'zepto', 'blinkit', 'swiggy', 'zomato', 'dunzo',
'decathlon', 'ikea', 'pepperfry', 'urban ladder'
],
'keywords': [
'order', 'shipped', 'delivered', 'delivery', 'tracking',
'purchase', 'cart', 'checkout', 'invoice', 'receipt',
'refund', 'return', 'exchange', 'dispatched', 'out for delivery'
],
},
EmailCategory.WORK: {
'senders': [
'linkedin', 'indeed', 'naukri', 'glassdoor', 'angel.co',
'slack', 'zoom', 'teams', 'meet', 'jira', 'confluence',
'github', 'gitlab', 'bitbucket', 'notion', 'asana', 'trello'
],
'keywords': [
'interview', 'meeting', 'agenda', 'project', 'deadline',
'review', 'standup', 'sprint', 'task', 'report', 'submission',
'application', 'resume', 'cv', 'job', 'position', 'hiring',
'salary', 'offer letter', 'joining', 'onboarding'
],
},
EmailCategory.NEWSLETTER: {
'senders': [
'substack', 'medium', 'morning brew', 'digest', 'newsletter',
'daily', 'weekly', 'update', 'news', 'times', 'hindu', 'express'
],
'keywords': [
'newsletter', 'digest', 'weekly', 'daily', 'update', 'news',
'article', 'read', 'story', 'trending', 'top stories',
'this week', 'this month', 'roundup', 'edition'
],
},
EmailCategory.PROMOTIONAL: {
'senders': [
'offer', 'deal', 'discount', 'sale', 'promo', 'marketing',
'shopify', 'mailchimp', 'campaign'
],
'keywords': [
'offer', 'discount', 'sale', 'deal', 'coupon', 'promo',
'limited time', 'exclusive', 'special', 'hurry', 'ends soon',
'flash sale', 'clearance', 'save', '% off', 'free shipping',
'buy now', 'shop now', 'don\'t miss'
],
},
EmailCategory.SOCIAL: {
'senders': [
'facebook', 'instagram', 'twitter', 'whatsapp', 'telegram',
'snapchat', 'tiktok', 'youtube', 'reddit', 'discord', 'quora'
],
'keywords': [
'friend request', 'like', 'comment', 'share', 'mentioned',
'tagged', 'message', 'follow', 'notification', 'birthday',
'invitation', 'event', 'rsvp', 'group'
],
},
}
# LLM prompt template
CLASSIFICATION_PROMPT: ClassVar[str] = """You are an email classifier. Analyze this email and categorize it.
EMAIL:
Subject: {subject}
From: {sender}
Body: {body}
TASK:
Classify this email into exactly ONE category.
CATEGORIES:
- finance: Banks, payments, transactions, investments, credit cards, loans
- shopping: Orders, deliveries, purchases, e-commerce, food delivery
- work: Job-related, recruitment, office, meetings, projects
- newsletter: Digests, subscriptions, blogs, articles
- promotional: Marketing, offers, discounts, advertisements
- social: Social networks, personal messages, invitations
- other: Anything that doesn't fit above
OUTPUT FORMAT (JSON only, no other text):
{{"category": "<category>", "confidence": "<high/medium/low>", "reason": "<brief reason>"}}
"""
def __init__(
self,
use_llm: bool = False,
model_path: Optional[str] = None,
debug: bool = False
) -> None:
"""
Initialize the EmailClassifier.
Args:
use_llm: If True, use LLM for classification (slower but more accurate).
model_path: Path to LLM model (required if use_llm=True).
debug: If True, enable debug logging.
Example:
>>> classifier = EmailClassifier() # Rule-based
>>> classifier = EmailClassifier(use_llm=True, model_path="path/to/model")
Raises:
ValueError: If use_llm=True but model_path not provided.
"""
self.use_llm = use_llm
self.debug = debug
self.model = None
self.tokenizer = None
if debug:
logger.setLevel(logging.DEBUG)
if use_llm:
if not model_path:
raise ValueError("model_path required when use_llm=True")
self._load_model(model_path)
logger.info(f"EmailClassifier initialized (use_llm={use_llm})")
def _load_model(self, model_path: str) -> None:
"""Load LLM model for classification."""
try:
from mlx_lm import load
self.model, self.tokenizer = load(model_path)
logger.info(f"Loaded LLM from {model_path}")
except ImportError:
logger.error("mlx_lm not installed. Install with: pip install mlx-lm")
raise
except Exception as e:
logger.error(f"Failed to load model: {e}")
raise
def classify(
self,
subject: str = "",
sender: str = "",
body: str = ""
) -> ClassificationResult:
"""
Classify an email into a category.
This method accepts the email components and returns a classification
result with category, confidence, and reasoning.
Args:
subject: Email subject line.
sender: Sender name or email address.
body: Email body text.
Returns:
ClassificationResult: Classification with category and confidence.
Example:
>>> classifier = EmailClassifier()
>>> result = classifier.classify(
... subject="Transaction Alert",
... sender="HDFC Bank",
... body="Rs.500 debited from your account"
... )
>>> print(result.category)
'finance'
>>> print(result.is_transaction)
True
Note:
At least one of subject, sender, or body should be non-empty.
Empty input returns 'other' category with low confidence.
"""
# Validate input
if not any([subject, sender, body]):
logger.warning("Empty input provided")
return ClassificationResult(
category=EmailCategory.OTHER.value,
confidence="low",
reason="No content to classify",
is_transaction=False
)
try:
if self.use_llm and self.model is not None:
return self._classify_llm(subject, sender, body)
else:
return self._classify_rules(subject, sender, body)
except Exception as e:
logger.error(f"Classification failed: {e}", exc_info=True)
return ClassificationResult(
category=EmailCategory.OTHER.value,
confidence="low",
reason=f"Classification error: {str(e)}",
is_transaction=False
)
def _classify_rules(
self,
subject: str,
sender: str,
body: str
) -> ClassificationResult:
"""Classify using rule-based approach."""
combined = f"{subject} {sender} {body}".lower()
# Check for transaction first
is_transaction = any(kw in combined for kw in self.TRANSACTION_KEYWORDS)
# Score each category
scores: Dict[EmailCategory, int] = {}
reasons: Dict[EmailCategory, List[str]] = {}
for category, patterns in self.CATEGORY_PATTERNS.items():
score = 0
matched = []
# Check sender patterns (strong signal)
sender_lower = sender.lower()
for pattern in patterns.get('senders', []):
if pattern in sender_lower:
score += 3
matched.append(f"sender:{pattern}")
# Check keyword patterns
for keyword in patterns.get('keywords', []):
if keyword in combined:
score += 1
if len(matched) < 3: # Limit reasons
matched.append(keyword)
if score > 0:
scores[category] = score
reasons[category] = matched
# Handle no matches
if not scores:
return ClassificationResult(
category=EmailCategory.OTHER.value,
confidence="low",
reason="No matching patterns found",
is_transaction=is_transaction
)
# Get highest scoring category
best_category = max(scores, key=lambda k: scores[k])
best_score = scores[best_category]
best_reasons = reasons.get(best_category, [])
# Determine confidence based on score
if best_score >= 5:
confidence = "high"
elif best_score >= 3:
confidence = "medium"
else:
confidence = "low"
reason = f"Matched: {', '.join(best_reasons[:3])}"
logger.debug(f"Classification: {best_category.value} ({confidence}), scores: {scores}")
return ClassificationResult(
category=best_category.value,
confidence=confidence,
reason=reason,
is_transaction=is_transaction,
scores={k.value: v for k, v in scores.items()} if self.debug else None
)
def _classify_llm(
self,
subject: str,
sender: str,
body: str
) -> ClassificationResult:
"""Classify using LLM."""
from mlx_lm import generate
# Truncate body if too long
max_body_length = 1000
body_truncated = body[:max_body_length] if len(body) > max_body_length else body
# Build prompt
prompt = self.CLASSIFICATION_PROMPT.format(
subject=subject,
sender=sender,
body=body_truncated
)
# Generate response
response = generate(
self.model,
self.tokenizer,
prompt=prompt,
max_tokens=100,
verbose=False
)
# Parse JSON response
return self._parse_llm_response(response)
def _parse_llm_response(self, response: str) -> ClassificationResult:
"""Parse LLM JSON response into ClassificationResult."""
# Find JSON in response
json_match = re.search(r'\{[^}]+\}', response, re.DOTALL)
if not json_match:
logger.warning(f"No JSON found in LLM response: {response[:100]}")
return ClassificationResult(
category=EmailCategory.OTHER.value,
confidence="low",
reason="Failed to parse LLM response",
is_transaction=False
)
try:
data = json.loads(json_match.group())
category = data.get("category", "other").lower()
# Validate category
if category not in [c.value for c in EmailCategory]:
category = "other"
return ClassificationResult(
category=category,
confidence=data.get("confidence", "medium"),
reason=data.get("reason", "LLM classification"),
is_transaction=category == "finance"
)
except json.JSONDecodeError as e:
logger.warning(f"JSON parse error: {e}")
return ClassificationResult(
category=EmailCategory.OTHER.value,
confidence="low",
reason="Invalid JSON in response",
is_transaction=False
)
def is_financial_email(self, subject: str, sender: str, body: str) -> bool:
"""
Quick check if email is financial.
Faster than full classification when you only need to know
if the email is finance-related.
Args:
subject: Email subject.
sender: Email sender.
body: Email body.
Returns:
bool: True if email is finance-related.
"""
result = self.classify(subject, sender, body)
return result.category == EmailCategory.FINANCE.value or result.is_transaction
# Convenience function
def classify_email(
subject: str = "",
sender: str = "",
body: str = ""
) -> ClassificationResult:
"""
Convenience function to classify an email without instantiating class.
Args:
subject: Email subject.
sender: Email sender.
body: Email body.
Returns:
ClassificationResult: Classification result.
Example:
>>> from src.data.classifier import classify_email
>>> result = classify_email(
... subject="Order Shipped",
... sender="Amazon",
... body="Your order is on the way"
... )
>>> print(result.category)
'shopping'
"""
return EmailClassifier().classify(subject, sender, body)
if __name__ == "__main__":
# Self-test
logging.basicConfig(level=logging.DEBUG)
classifier = EmailClassifier(debug=True)
test_cases = [
("Transaction Alert", "HDFC Bank", "Rs.500 debited from your account"),
("Your order has shipped", "Amazon.in", "Order #123 is on the way"),
("Interview Invitation", "LinkedIn", "You have an interview scheduled"),
("Weekly Digest", "Substack", "Top 10 articles this week"),
]
for subject, sender, body in test_cases:
result = classifier.classify(subject, sender, body)
print(f"\n{subject} | {sender}")
print(f" → {result.category} ({result.confidence})")
print(f" → {result.reason}")
print(f" → Transaction: {result.is_transaction}")
|