spacy_matcher.py

import spacy
from spacy.matcher import Matcher


nlp = spacy.load("en_core_web_sm")
matcher = Matcher(nlp.vocab)


clause_patterns = {
    "CONFIDENTIALITY": [
        [{"LOWER": "confidentiality"}],
        [{"LOWER": "non-disclosure"}],
        [{"LOWER": "nda"}],
        [{"LOWER": "proprietary"}, {"LOWER": "information"}],
    ],
    "TERMINATION": [
        [{"LOWER": "termination"}],
        [{"LOWER": "end"}, {"LOWER": "of"}, {"LOWER": "agreement"}],
        [{"LOWER": "terminate"}, {"LOWER": "this"}, {"LOWER": "agreement"}],
    ],
    "NON_COMPETE": [
        [{"LOWER": "non-compete"}],
        [{"LOWER": "non"}, {"LOWER": "compete"}],
        [{"LOWER": "competition"}, {"LOWER": "restriction"}],
    ],
    "GOVERNING_LAW": [
        [{"LOWER": "governing"}, {"LOWER": "law"}],
        [{"LOWER": "jurisdiction"}],
        [{"LOWER": "choice"}, {"LOWER": "of"}, {"LOWER": "law"}],
    ],
    "SEVERABILITY": [
        [{"LOWER": "severability"}],
        [{"LOWER": "invalidity"}, {"OP": "?"}, {"LOWER": "provision"}],
        [{"LOWER": "severable"}],
    ],
    "LIABILITY": [
        [{"LOWER": "liability"}],
        [{"LOWER": "limitation"}, {"LOWER": "of"}, {"LOWER": "liability"}],
        [{"LOWER": "indemnification"}],
        [{"LOWER": "cap"}, {"LOWER": "on"}, {"LOWER": "damages"}],
    ],
    "FORCE_MAJEURE": [
        [{"LOWER": "force"}, {"LOWER": "majeure"}],
        [{"LOWER": "acts"}, {"LOWER": "of"}, {"LOWER": "god"}],
        [{"LOWER": "unforeseen"}, {"LOWER": "events"}],
        [{"LOWER": "pandemic"}],
    ],
    "PAYMENT_TERMS": [
        [{"LOWER": "payment"}, {"LOWER": "terms"}],
        [{"LOWER": "due"}, {"LOWER": "within"}, {"IS_DIGIT": True}, {"LOWER": {"IN": ["days", "weeks"]}}],
        [{"LOWER": "invoice"}, {"LOWER": "shall"}, {"LOWER": "be"}],
        [{"LOWER": "net"}, {"IS_DIGIT": True}],
    ]
}

# Adding all patterns to matcher
for label, patterns in clause_patterns.items():
    for pattern in patterns:
        matcher.add(label, [pattern])

print("Clause matcher loaded with extended patterns.")

def find_clauses(text, window_size=30):
    """Find clauses in text and return context."""
    doc = nlp(text)
    matches = matcher(doc)
    results = []
    for match_id, start, end in matches:
        span = doc[start:end]
        label = nlp.vocab.strings[match_id]
        # Get context window
        ctx_start = max(0, start - window_size)
        ctx_end = min(len(doc), end + window_size)
        context = doc[ctx_start:ctx_end].text
        results.append((label, context, start, end))
    return results