flask-ocr-api / app /parser.py.bak
vamsi1103's picture
Upload 26 files
0e157b6 verified
Raw History Blame Contribute Delete
4.19 kB
import re
from datetime import datetime
def parse_ocr_results(ocr_data):
"""
Parses raw OCR extracted text into a structured JSON format with items and rates.
ocr_data is a list of dicts: [{"text": "Tomato", "confidence": 0.98}, ...]
"""
items = []
date_str = datetime.now().strftime("%Y-%m-%d")
if not ocr_data:
return {
"success": True,
"date": date_str,
"items": [],
"debug_raw_text": []
}
current_item = None
for data in ocr_data:
text = data["text"].strip()
conf = data["confidence"]
# Try to find a date in the text
date_match = re.search(r'(\d{2,4}[-/]\d{1,2}[-/]\d{2,4})', text)
if date_match:
date_str = date_match.group(1)
continue
# Ignore purely serial numbers (e.g. just "1", "2")
if re.match(r'^\d+$', text):
continue
# Check if item and rate are on the same line (e.g. "Tomato 40.00", "Tomato40.00", or "వంకాయలు 40.00")
same_line_match = re.search(r'^(.*?[^\d\s/])\s*([\d\./]+)$', text)
if same_line_match:
name = same_line_match.group(1).strip()
rate_text = same_line_match.group(2).strip()
# Remove leading serial numbers from name (e.g. "1. Tomato" -> "Tomato")
name = re.sub(r'^\d+[\.\-\s]+', '', name).strip()
# Ensure name has some non-digit characters
if re.search(r'[^\d\.\s/]', name):
_add_item(items, name, rate_text, conf)
current_item = None
continue
# Check if it's purely a rate (e.g., "40.00" or "100/150")
rate_match = re.match(r'^([\d\.]+)(?:/([\d\.]+))?$', text.replace(" ", ""))
if rate_match and current_item:
rate_text = text
_add_item(items, current_item["name"], rate_text, min(current_item["confidence"], conf))
current_item = None
continue
# Otherwise, if it has letters/non-digits, treat it as an item name
if re.search(r'[^\d\.\s/]', text):
# Ignore known header words that appear before the table starts
if any(h in text for h in ["రైతుబజార", "ధరలు", "కూరగాయలురకం", "కేజ", "ధర", "సంఖత"]):
current_item = None
continue
# Remove leading serial numbers
clean_name = re.sub(r'^\d+[\.\-\s]+', '', text).strip()
if clean_name:
if current_item:
# If the name is getting suspiciously long (like a header), reset it!
if len(current_item["name"]) > 40:
current_item = {"name": clean_name, "confidence": conf}
else:
current_item["name"] += " " + clean_name
current_item["confidence"] = min(current_item["confidence"], conf)
else:
current_item = {"name": clean_name, "confidence": conf}
return {
"success": True,
"date": date_str,
"items": items,
"debug_raw_text": [d["text"] for d in ocr_data]
}
def _add_item(items_list, name, rate_text, conf):
clean_rate = rate_text.replace(" ", "")
if "/" in clean_rate:
try:
rates = [float(x) for x in clean_rate.split("/") if x]
items_list.append({
"name": name,
"rate_type": "multiple",
"rate": None,
"available_rates": rates,
"rate_text": rate_text,
"confidence": round(conf, 2)
})
except ValueError:
pass
else:
try:
rate_val = float(clean_rate)
items_list.append({
"name": name,
"rate_type": "single",
"rate": rate_val,
"rate_text": rate_text,
"confidence": round(conf, 2)
})
except ValueError:
pass