contract-clause-extractor

Contract Clause Extractor

Overview

Extract and analyze key clauses from construction contracts using NLP. Identify critical provisions for payment, changes, disputes, warranties, and risk allocation. Support contract review and compliance tracking.

Key Clause Categories

┌─────────────────────────────────────────────────────────────────┐
│                 CONTRACT CLAUSE CATEGORIES                       │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  Payment           Changes            Risk                       │
│  ───────           ───────            ────                       │
│  📅 Pay schedule   📝 CO process      ⚠️ Indemnification        │
│  💰 Retainage      ⏰ Notice period   🔒 Insurance              │
│  📋 Requirements   💵 Pricing         🏛️ Liability limits       │
│                                                                  │
│  Schedule          Disputes           Closeout                   │
│  ────────          ────────           ────────                   │
│  📆 Milestones     ⚖️ Resolution      ✅ Punch list             │
│  💸 Liquidated $   🏛️ Jurisdiction    📄 Warranties             │
│  ⏱️ Extensions     👤 Mediation       🔑 Final payment          │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Technical Implementation

from dataclasses import dataclass, field
from typing import List, Dict, Optional, Tuple
from enum import Enum
import re

class ClauseCategory(Enum):
    PAYMENT = "payment"
    CHANGE_ORDER = "change_order"
    SCHEDULE = "schedule"
    DISPUTE = "dispute"
    INSURANCE = "insurance"
    WARRANTY = "warranty"
    TERMINATION = "termination"
    INDEMNIFICATION = "indemnification"
    SAFETY = "safety"
    CLOSEOUT = "closeout"

class RiskLevel(Enum):
    HIGH = "high"
    MEDIUM = "medium"
    LOW = "low"

@dataclass
class ExtractedClause:
    category: ClauseCategory
    article_number: str
    title: str
    full_text: str
    key_terms: List[str]
    dollar_amounts: List[float]
    time_periods: List[str]
    risk_level: RiskLevel
    notes: str = ""

@dataclass
class ContractSummary:
    contract_type: str
    parties: Dict[str, str]
    contract_value: float
    duration_days: int
    start_date: str
    key_dates: Dict[str, str]
    clauses_by_category: Dict[str, List[ExtractedClause]]
    risk_assessment: Dict[str, str]
    action_items: List[str]

class ContractClauseExtractor:
    """Extract key clauses from construction contracts."""

    # Patterns for clause identification
    CLAUSE_PATTERNS = {
        ClauseCategory.PAYMENT: [
            r"payment\s+(terms|schedule|application)",
            r"progress\s+payment",
            r"retainage|retention",
            r"pay\s+when\s+paid",
            r"pay\s+if\s+paid",
            r"final\s+payment",
        ],
        ClauseCategory.CHANGE_ORDER: [
            r"change\s+order",
            r"change\s+directive",
            r"modifications?\s+to\s+contract",
            r"extra\s+work",
            r"changes\s+in\s+the\s+work",
        ],
        ClauseCategory.SCHEDULE: [
            r"time\s+of\s+completion",
            r"liquidated\s+damages",
            r"delay|extension\s+of\s+time",
            r"substantial\s+completion",
            r"schedule\s+of\s+values",
            r"milestone",
        ],
        ClauseCategory.DISPUTE: [
            r"dispute\s+resolution",
            r"mediation",
            r"arbitration",
            r"claims?\s+procedures?",
            r"litigation",
            r"governing\s+law",
        ],
        ClauseCategory.INSURANCE: [
            r"insurance\s+requirements?",
            r"builder'?s?\s+risk",
            r"general\s+liability",
            r"professional\s+liability",
            r"workers'?\s+compensation",
        ],
        ClauseCategory.WARRANTY: [
            r"warranty|warranties",
            r"guarantee",
            r"correction\s+of\s+work",
            r"defect",
        ],
        ClauseCategory.INDEMNIFICATION: [
            r"indemnif",
            r"hold\s+harmless",
            r"defend",
            r"limitation\s+of\s+liability",
        ],
        ClauseCategory.TERMINATION: [
            r"termination",
            r"suspension\s+of\s+work",
            r"default",
            r"for\s+cause|for\s+convenience",
        ],
    }

    # Key terms to extract
    KEY_TERM_PATTERNS = {
        "dollar_amounts": r'\$[\d,]+(?:\.\d{2})?|\d+(?:,\d{3})*\s*dollars',
        "percentages": r'\d+(?:\.\d+)?%',
        "time_periods": r'\d+\s*(?:days?|weeks?|months?|years?|business\s+days?|calendar\s+days?)',
        "notice_periods": r'(?:within|not\s+(?:less|more)\s+than)\s+\d+\s*(?:days?|hours?)',
    }

    def __init__(self):
        self.extracted_clauses: List[ExtractedClause] = []

    def extract_clauses(self, contract_text: str) -> List[ExtractedClause]:
        """Extract all identifiable clauses from contract text."""
        self.extracted_clauses = []

        # Split into articles/sections
        sections = self._split_into_sections(contract_text)

        for section_num, section_text in sections.items():
            # Identify clause category
            category = self._identify_category(section_text)

            if category:
                clause = self._extract_clause_details(
                    category, section_num, section_text
                )
                self.extracted_clauses.append(clause)

        return self.extracted_clauses

    def _split_into_sections(self, text: str) -> Dict[str, str]:
        """Split contract into numbered sections."""
        sections = {}

        # Pattern for article/section headers
        header_pattern = r'(?:ARTICLE|SECTION|Article|Section)\s+(\d+(?:\.\d+)?)[:\.]?\s*([A-Z][A-Za-z\s]+)?'

        parts = re.split(header_pattern, text)

        current_num = "0"
        current_text = ""

        for i, part in enumerate(parts):
            if re.match(r'^\d+(?:\.\d+)?$', part.strip()):
                if current_text:
                    sections[current_num] = current_text
                current_num = part.strip()
                current_text = ""
            else:
                current_text += part

        if current_text:
            sections[current_num] = current_text

        return sections

    def _identify_category(self, text: str) -> Optional[ClauseCategory]:
        """Identify clause category from text."""
        text_lower = text.lower()

        for category, patterns in self.CLAUSE_PATTERNS.items():
            for pattern in patterns:
                if re.search(pattern, text_lower):
                    return category

        return None

    def _extract_clause_details(self, category: ClauseCategory,
                               section_num: str, text: str) -> ExtractedClause:
        """Extract detailed information from clause."""
        # Extract title (first line or capitalized phrase)
        title_match = re.search(r'^([A-Z][A-Z\s]+)', text.strip())
        title = title_match.group(1).strip() if title_match else f"Section {section_num}"

        # Extract dollar amounts
        dollar_amounts = []
        for match in re.finditer(self.KEY_TERM_PATTERNS["dollar_amounts"], text):
            amount_str = match.group().replace('$', '').replace(',', '').replace('dollars', '').strip()
            try:
                dollar_amounts.append(float(amount_str))
            except ValueError:
                pass

        # Extract time periods
        time_periods = re.findall(self.KEY_TERM_PATTERNS["time_periods"], text, re.IGNORECASE)

        # Extract key terms
        key_terms = self._extract_key_terms(category, text)

        # Assess risk level
        risk_level = self._assess_risk(category, text)

        return ExtractedClause(
            category=category,