dataSraping / data_normalizer.py
Ai-Bharti's picture
Upload 2 files
e8d8f32 verified
Raw History Blame Contribute Delete
19.6 kB
"""
data_normalizer.py - Smart Data Extraction + Normalization
===========================================================
Yeh module kya karta hai:
βœ… USD β†’ INR Cr convert karta hai
βœ… Million/Billion/Lakh β†’ Cr convert karta hai
βœ… Timeline ko standard format mein laata hai
βœ… Customer, Order Type, Date accurately nikalta hai
βœ… SEBI Annexure-A table structure samajhta hai (jo aapke PDFs mein hai)
Real-world examples jo handle karta hai:
USD 1,346,210 β†’ 1.14 Cr (@ Rs.85/USD)
β‚Ή1,587.10 million β†’ 158.71 Cr
Rs. 890.69 Million β†’ 89.07 Cr
Rs. 17,750 Million β†’ 1775.00 Cr
INR 20.58 Crores β†’ 20.58 Cr
Rs. 1,827.33 β†’ 1827.33 Cr (already in Cr)
Delivery by Oct 2027 β†’ Oct 2027
Within 8 months β†’ 8 Months
910 Days β†’ 910 Days (~30 Months)
4 (04) Years β†’ 4 Years (~48 Months)
18 Months β†’ 18 Months
"""
import re
# ─────────────────────────────────────────────────────────
# πŸ’± CURRENCY CONVERSION RATES
# Inhe update karte rahna agar rate badal jaye
# ─────────────────────────────────────────────────────────
USD_TO_INR = 85.0 # 1 USD = Rs. 85 approx
MILLION_TO_CR = 0.1 # 1 Million = 0.1 Cr (10 Lakh)
BILLION_TO_CR = 100.0 # 1 Billion = 100 Cr
LAKH_TO_CR = 0.01 # 1 Lakh = 0.01 Cr
MN_TO_CR = 0.1 # MN = Million shorthand
# ════════════════════════════════════════════════════════
# πŸ’° MAIN FUNCTION: Order Value Normalize karo
# ════════════════════════════════════════════════════════
def normalize_order_value(text: str) -> dict:
"""
Kisi bhi format ki value ko INR Crore mein convert karta hai.
Input examples:
"USD 1,346,210"
"β‚Ή1,587.10 million (Including GST)"
"Rs. 890.69 Million"
"Rs. 1,827.33 Cr"
"INR 20.58 Crores"
"Rs. 17,750 Million"
Return:
{
"original": "USD 1,346,210",
"inr_crore": 1.14,
"inr_crore_display": "1.14 Cr",
"currency": "USD",
"note": "Converted @ Rs.85/USD"
}
"""
if not text or text.strip() in ["N/A", "", "-", "β€”"]:
return _empty_value()
clean = text.strip()
# ── Pattern 1: USD amount ──
# Matches: "USD 1,346,210" or "USD 1346210" or "US$ 1,346,210"
usd_match = re.search(
r"USD\s*[\$]?\s*([\d,]+(?:\.\d+)?)",
clean, re.IGNORECASE
)
if usd_match:
usd_val = float(usd_match.group(1).replace(",", ""))
inr_val = usd_val * USD_TO_INR
inr_cr = round(inr_val / 1e7, 2) # Convert to Crore
return {
"original": clean,
"inr_crore": inr_cr,
"inr_crore_display": f"{inr_cr:.2f} Cr",
"currency": "USD",
"note": f"USD {usd_val:,.0f} Γ— Rs.{USD_TO_INR}/USD Γ· 1Cr"
}
# ── Pattern 2: Rupees in Million ──
# Matches: "Rs. 890.69 Million", "β‚Ή1,587.10 million", "INR 17,750 Million", "Rs. 17,750 Mn"
million_match = re.search(
r"(?:Rs\.?|β‚Ή|INR|INR\s*)?\s*([\d,]+(?:\.\d+)?)\s*(?:million|mn\.?)\b",
clean, re.IGNORECASE
)
if million_match:
mn_val = float(million_match.group(1).replace(",", ""))
cr_val = round(mn_val * MILLION_TO_CR, 2)
return {
"original": clean,
"inr_crore": cr_val,
"inr_crore_display": f"{cr_val:.2f} Cr",
"currency": "INR",
"note": f"Rs.{mn_val} Mn Γ· 10 = {cr_val} Cr"
}
# ── Pattern 3: Rupees in Billion ──
# Matches: "Rs. 2.5 Billion", "INR 1 Billion"
billion_match = re.search(
r"(?:Rs\.?|β‚Ή|INR)?\s*([\d,]+(?:\.\d+)?)\s*(?:billion|bn\.?)\b",
clean, re.IGNORECASE
)
if billion_match:
bn_val = float(billion_match.group(1).replace(",", ""))
cr_val = round(bn_val * BILLION_TO_CR, 2)
return {
"original": clean,
"inr_crore": cr_val,
"inr_crore_display": f"{cr_val:.2f} Cr",
"currency": "INR",
"note": f"Rs.{bn_val} Bn Γ— 100 = {cr_val} Cr"
}
# ── Pattern 4: Already in Crore ──
# Matches: "Rs. 1,827.33 Cr", "INR 20.58 Crores", "Rs. 775.27 Cr. approx"
# Also: "Rs 148 crores", "approx. Rs. 33 Crore"
crore_match = re.search(
r"(?:Rs\.?|β‚Ή|INR)?\s*([\d,]+(?:\.\d+)?)\s*(?:crore|cr\.?|crores)\b",
clean, re.IGNORECASE
)
if crore_match:
cr_val = round(float(crore_match.group(1).replace(",", "")), 2)
return {
"original": clean,
"inr_crore": cr_val,
"inr_crore_display": f"{cr_val:.2f} Cr",
"currency": "INR",
"note": "Already in Crore"
}
# ── Pattern 5: Rupees in Lakh ──
# Matches: "Rs. 50 Lakh", "Rs. 2.5 L"
lakh_match = re.search(
r"(?:Rs\.?|β‚Ή|INR)?\s*([\d,]+(?:\.\d+)?)\s*(?:lakh|lac|l)\b",
clean, re.IGNORECASE
)
if lakh_match:
lakh_val = float(lakh_match.group(1).replace(",", ""))
cr_val = round(lakh_val * LAKH_TO_CR, 4)
return {
"original": clean,
"inr_crore": cr_val,
"inr_crore_display": f"{cr_val:.4f} Cr",
"currency": "INR",
"note": f"Rs.{lakh_val} L Γ· 100 = {cr_val} Cr"
}
# ── Fallback: Kuch match nahi hua ──
return {
"original": clean,
"inr_crore": None,
"inr_crore_display": clean, # As-is rakh do
"currency": "UNKNOWN",
"note": "Auto-convert nahi hua, manual check karo"
}
def _empty_value():
return {
"original": "", "inr_crore": None,
"inr_crore_display": "N/A", "currency": "", "note": ""
}
# ════════════════════════════════════════════════════════
# ⏰ TIMELINE NORMALIZE karo
# ════════════════════════════════════════════════════════
def normalize_timeline(text: str) -> dict:
"""
Kisi bhi format ki timeline ko standard format mein convert karta hai.
Input examples:
"910 Days" β†’ {"display": "910 Days (~30 Months)", "months": 30}
"4 (04) Years" β†’ {"display": "4 Years (~48 Months)", "months": 48}
"Within 18 Months" β†’ {"display": "18 Months", "months": 18}
"Delivery to be completed by Aug 31, 2026" β†’ {"display": "Aug 2026", "months": None}
"Delivery by October 2027" β†’ {"display": "Oct 2027", "months": None}
"Within 8 months from LOA" β†’ {"display": "8 Months from LOA", "months": 8}
"11 months from Letter of Acceptance" β†’ same pattern
"""
if not text or text.strip() in ["N/A", "", "-"]:
return {"display": "N/A", "months": None, "original": text}
clean = text.strip()
months = None
display = clean # Default: as-is
# ── Pattern 1: Days ──
# "910 Days", "30 days"
days_match = re.search(r"(\d+)\s*days?\b", clean, re.IGNORECASE)
if days_match:
days = int(days_match.group(1))
months = round(days / 30.4)
display = f"{days} Days (~{months} Months)"
return {"display": display, "months": months, "original": clean}
# ── Pattern 2: Years ──
# "4 Years", "4 (04) Years", "Four (04) Years", "2.5 years"
years_match = re.search(
r"(\d+(?:\.\d+)?)\s*(?:\(\d+\)\s*)?years?\b",
clean, re.IGNORECASE
)
if years_match:
years = float(years_match.group(1))
months = round(years * 12)
display = f"{int(years) if years == int(years) else years} Years (~{months} Months)"
return {"display": display, "months": months, "original": clean}
# ── Pattern 3: Months ──
# "18 Months", "Within 18 Months", "8 months from LOA"
months_match = re.search(r"(\d+)\s*months?\b", clean, re.IGNORECASE)
if months_match:
months = int(months_match.group(1))
# Context nikalo - "from LOA" jaisi cheez
context_match = re.search(
r"months?\s+(from\s+[^\.,]{3,40})", clean, re.IGNORECASE
)
suffix = f" {context_match.group(1)}" if context_match else ""
display = f"{months} Months{suffix}"
return {"display": display, "months": months, "original": clean}
# ── Pattern 4: Specific Date ──
# "Delivery by October 2027", "by August 31, 2026", "by April 2026"
date_match = re.search(
r"(?:by|until|before|on)\s+"
r"((?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]*"
r"(?:\s+\d{1,2},?)?\s+\d{4})",
clean, re.IGNORECASE
)
if date_match:
display = date_match.group(1).strip()
# Short month format: "October 2027" β†’ "Oct 2027"
display = _shorten_month(display)
return {"display": display, "months": None, "original": clean}
# ── Fallback ──
return {"display": clean[:60], "months": None, "original": clean}
def _shorten_month(date_str: str) -> str:
"""'October 2027' β†’ 'Oct 2027'"""
month_map = {
"january": "Jan", "february": "Feb", "march": "Mar",
"april": "Apr", "may": "May", "june": "Jun",
"july": "Jul", "august": "Aug", "september": "Sep",
"october": "Oct", "november": "Nov", "december": "Dec"
}
for full, short in month_map.items():
date_str = re.sub(full, short, date_str, flags=re.IGNORECASE)
return date_str
# ════════════════════════════════════════════════════════
# πŸ“‹ SEBI ANNEXURE-A TABLE PARSER
# Yeh special function hai jo BSE filings ki standard table padh sakta hai
# ════════════════════════════════════════════════════════
def parse_sebi_annexure_table(text: str) -> dict:
"""
SEBI Regulation 30 ke standard Annexure-A table se data nikalta hai.
Yeh table structure handle karta hai:
Row 1: Name of entity awarding β†’ Customer
Row 2: Significant terms β†’ Order Description
Row 4: Nature of order β†’ Order Type
Row 6: Time period β†’ Timeline
Row 7: Broad consideration / size β†’ Order Value
Yeh almost SABB BSE filings mein hota hai!
"""
result = {
"customer": "",
"order_description": "",
"order_type": "",
"timeline_raw": "",
"value_raw": "",
"domestic_or_intl": "",
}
if not text:
return result
# Normalize whitespace
text_clean = re.sub(r"\s+", " ", text)
# ── Row 1: Customer/Entity Name ──
cust_patterns = [
r"(?:name of the entity awarding[^:]*?[:]\s*)([^\n\r]{5,100}?)(?:\n|\r|$|\d+\.)",
r"(?:entity awarding[^:]*?[:]\s*)([^\n\r]{5,100}?)(?:\n|\r|$)",
r"(?:awarded by|awarded from)[:\s]+([A-Z][^\n\r]{5,100}?)(?:\n|\r|$)",
]
for p in cust_patterns:
m = re.search(p, text_clean, re.IGNORECASE)
if m:
val = m.group(1).strip()
# Tabular "Details" header skip karo
if len(val) > 3 and "particular" not in val.lower():
result["customer"] = val[:120]
break
# ── Row 2: Order Description / Terms ──
desc_patterns = [
r"(?:significant terms[^:]*?[:]\s*)([^\n\r]{10,300}?)(?:\n|\r|$|\d+\.)",
r"(?:terms and conditions[^:]*?[:]\s*)([^\n\r]{10,300}?)(?:\n|\r|$)",
]
for p in desc_patterns:
m = re.search(p, text_clean, re.IGNORECASE)
if m:
result["order_description"] = m.group(1).strip()[:300]
break
# ── Row 4: Nature of Order ──
nature_patterns = [
r"(?:nature of order[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$|\d+\.)",
]
for p in nature_patterns:
m = re.search(p, text_clean, re.IGNORECASE)
if m:
result["order_type"] = m.group(1).strip()[:150]
break
# ── Domestic/International ──
dom_match = re.search(
r"(?:domestic or international[^:]*?[:]\s*)([^\n\r]{3,30}?)(?:\n|\r|$|\d+\.)",
text_clean, re.IGNORECASE
)
if dom_match:
result["domestic_or_intl"] = dom_match.group(1).strip()
# ── Row 6: Timeline ──
time_patterns = [
r"(?:time period[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$|\d+\.)",
r"(?:period by which[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$)",
r"(?:to be executed[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$)",
]
for p in time_patterns:
m = re.search(p, text_clean, re.IGNORECASE)
if m:
result["timeline_raw"] = m.group(1).strip()[:200]
break
# ── Row 7: Order Value ──
value_patterns = [
r"(?:broad consideration[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$|\d+\.)",
r"(?:size of the order[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$)",
r"(?:consideration or size[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$)",
r"(?:value of order[^:]*?[:]\s*)([^\n\r]{5,200}?)(?:\n|\r|$)",
]
for p in value_patterns:
m = re.search(p, text_clean, re.IGNORECASE)
if m:
result["value_raw"] = m.group(1).strip()[:200]
break
return result
# ════════════════════════════════════════════════════════
# 🎯 MASTER FUNCTION - Sab ek jagah
# ════════════════════════════════════════════════════════
def extract_and_normalize(pdf_text: str, snippet: str, filing_meta: dict) -> dict:
"""
PDF text se FULL normalized record banata hai.
Pehle SEBI Annexure table parse karta hai (most reliable),
phir value + timeline normalize karta hai.
Return: Complete record dict for CSV
"""
# Step 1: SEBI standard table parse karo
annexure = parse_sebi_annexure_table(pdf_text)
# Step 2: Value normalize karo
raw_value = annexure["value_raw"] or snippet
value_info = normalize_order_value(raw_value)
# Step 3: Timeline normalize karo
raw_timeline = annexure["timeline_raw"] or ""
timeline_info = normalize_timeline(raw_timeline)
# Step 4: Customer finalize
customer = annexure["customer"]
if not customer:
# Fallback: snippet se try karo
m = re.search(
r"(?:from|by)\s+([A-Z][A-Za-z\s&\-]{5,60}"
r"(?:Ltd\.?|Limited|Authority|Board|NHAI|DDA|NTPC|ONGC|RAIL|NAVY|ARMY))",
snippet
)
if m:
customer = m.group(1).strip()
# Step 5: Order type finalize
order_type = annexure["order_type"] or filing_meta.get("title", "")
# Title se Order type nikalo agar table mein nahi mila
if "LOA" in order_type or "Letter of Award" in (pdf_text + snippet):
order_type = "Letter of Award"
elif "LOI" in order_type or "Letter of Intent" in (pdf_text + snippet):
order_type = "Letter of Intent"
elif "Work Order" in (pdf_text + snippet):
order_type = "Work Order"
return {
"Stock Name": filing_meta.get("stock", ""),
"Exchange": filing_meta.get("exchange", ""),
"Filing Title": filing_meta.get("title", ""),
"Customer": customer or "N/A",
"Order Description": annexure["order_description"] or "N/A",
"Order Type": order_type or "N/A",
"Domestic / International": annexure["domestic_or_intl"] or "N/A",
"Date Order Received": filing_meta.get("date", ""),
# βœ… NORMALIZED VALUES
"Order Value (INR Cr)": value_info["inr_crore_display"],
"Value (Raw)": value_info["original"],
"Conversion Note": value_info["note"],
# βœ… NORMALIZED TIMELINE
"Timeline": timeline_info["display"],
"Timeline (Months)": str(timeline_info["months"]) if timeline_info["months"] else "N/A",
"Timeline (Raw)": timeline_info["original"],
"Source PDF": filing_meta.get("pdf_url", ""),
"Screener Date": filing_meta.get("date", ""),
"Scraped At": __import__("datetime").datetime.now().strftime("%Y-%m-%d %H:%M"),
}
# ════════════════════════════════════════════════════════
# πŸ§ͺ TEST - Seedha run karo to verify
# ════════════════════════════════════════════════════════
if __name__ == "__main__":
print("=" * 55)
print("πŸ’° ORDER VALUE CONVERSION TESTS")
print("=" * 55)
test_values = [
"USD 1,346,210",
"USD 1346210.00",
"β‚Ή1,587.10 million (Including GST)",
"Rs. 890.69 Million (including GST)",
"Rs. 17,750 Million",
"INR 20.58 Crores (Including Tax)",
"Rs. 1,827.33 Cr",
"Rs. 148 crores",
"Rs. 775.27 Cr. approx.",
"Rs. 33 Crore (Excl. GST)",
"Rs. 2.13 crores (excluding taxes)",
"Rs. 437.79 Cr. (Approx.)",
"Rs. 1,582.11", # Pure number - Cr assume karein
]
for val in test_values:
result = normalize_order_value(val)
print(f" Input : {val}")
print(f" Output: {result['inr_crore_display']} [{result['note']}]")
print()
print("=" * 55)
print("⏰ TIMELINE NORMALIZATION TESTS")
print("=" * 55)
test_timelines = [
"910 Days",
"Four (04) Years",
"4 (04) Years",
"Within 18 Months from signing of PPA",
"Delivery to be completed by August 31, 2026",
"Delivery by October 2027",
"Within 8 months from the Letter of Award",
"11 months from the date of issue of Letter of Acceptance",
"Delivery to be completed by April 2026",
"25 years from commissioning",
]
for tl in test_timelines:
result = normalize_timeline(tl)
months_str = f" | {result['months']} months" if result["months"] else ""
print(f" Input : {tl}")
print(f" Output: {result['display']}{months_str}")
print()