Datasets:
File size: 1,721 Bytes
2fdc260 | 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 | # Hitung akurasi jawaban (\boxed{...} vs gtruth_answer). Contoh:
# python scripts/evaluate_json.py --json_path data/pajak-eval.json --gtruth_answer gtruth_answer --generated_text message_content
import argparse, json, re
def extract_boxed(text):
i = text.rfind("\\boxed{")
if i < 0: return None
j, depth, out = i + 7, 1, []
while j < len(text) and depth:
c = text[j]
if c == "{": depth += 1
elif c == "}": depth -= 1
if depth: out.append(c)
j += 1
return "".join(out) if depth == 0 else None
def to_num(s):
m = re.search(r"-?\d[\d.,]*", s.replace("\\,", ""))
if not m: return None
g = m.group(0).rstrip(".,")
if re.fullmatch(r"\d{1,3}(\.\d{3})+(,\d+)?", g): g = g.replace(".", "").replace(",", ".")
elif re.fullmatch(r"\d{1,3}(,\d{3})+(\.\d+)?", g): g = g.replace(",", "")
else: g = g.replace(",", ".")
try: return float(g)
except ValueError: return None
def norm(s): return re.sub(r"\W+", " ", s.lower()).strip()
parser = argparse.ArgumentParser()
parser.add_argument("--json_path", required=True)
parser.add_argument("--gtruth_answer", default="gtruth_answer")
parser.add_argument("--generated_text", default="message_content")
args = parser.parse_args()
data = json.load(open(args.json_path, encoding="utf-8"))
correct = 0
for d in data:
pred, gt = extract_boxed(d[args.generated_text]), d[args.gtruth_answer]
if pred is None: continue
a, b = to_num(gt), to_num(pred)
if a is not None and b is not None: correct += abs(a - b) < 0.5
else: correct += norm(gt) == norm(pred) or norm(gt) in norm(pred) or norm(pred) in norm(gt)
print(f"Answer accuracy: {correct/len(data):.1%} ({correct}/{len(data)})")
|