Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -143,6 +143,8 @@ def run_ner(text: str):
|
|
| 143 |
|
| 144 |
rows_html = ""
|
| 145 |
row_num = 1
|
|
|
|
|
|
|
| 146 |
for ent in results:
|
| 147 |
raw_label = ent["entity_group"]
|
| 148 |
label_id = get_label_id(raw_label)
|
|
@@ -151,6 +153,12 @@ def run_ner(text: str):
|
|
| 151 |
if not word:
|
| 152 |
continue
|
| 153 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
hex_color = LABEL_HEX.get(label_id, "#e2e8f0")
|
| 155 |
score = f"{ent['score']:.2%}"
|
| 156 |
even = row_num % 2 == 0
|
|
@@ -171,6 +179,9 @@ def run_ner(text: str):
|
|
| 171 |
</tr>"""
|
| 172 |
row_num += 1
|
| 173 |
|
|
|
|
|
|
|
|
|
|
| 174 |
table_html = f"""
|
| 175 |
<div style="overflow-x:auto; margin-top:4px; border-radius:12px;
|
| 176 |
border:1px solid #e2e8f0; box-shadow:0 2px 12px rgba(0,0,0,0.06);">
|
|
@@ -193,12 +204,8 @@ def run_ner(text: str):
|
|
| 193 |
</div>"""
|
| 194 |
return table_html
|
| 195 |
|
| 196 |
-
|
| 197 |
# PDF NER
|
| 198 |
def run_ner_pdf(pdf_file):
|
| 199 |
-
"""Highlight NER entities in PDF and return (output_path, legend_html, warning_html)."""
|
| 200 |
-
|
| 201 |
-
# No file uploaded
|
| 202 |
if pdf_file is None:
|
| 203 |
return (
|
| 204 |
None,
|
|
@@ -207,11 +214,9 @@ def run_ner_pdf(pdf_file):
|
|
| 207 |
)
|
| 208 |
|
| 209 |
pdf_path = pdf_file if isinstance(pdf_file, str) else pdf_file.name
|
| 210 |
-
|
| 211 |
doc = fitz.open(pdf_path)
|
| 212 |
page_count = len(doc)
|
| 213 |
|
| 214 |
-
# Page limit check
|
| 215 |
if page_count > MAX_PDF_PAGES:
|
| 216 |
doc.close()
|
| 217 |
warn_html = f"""
|
|
@@ -224,58 +229,81 @@ def run_ner_pdf(pdf_file):
|
|
| 224 |
<p style="margin:4px 0 0; color:#9f1239; font-size:13px;">
|
| 225 |
File Anda memiliki <strong>{page_count} halaman</strong>.
|
| 226 |
Maksimal yang diizinkan adalah <strong>{MAX_PDF_PAGES} halaman</strong>.
|
| 227 |
-
Silakan unggah PDF yang lebih pendek atau potong halamannya terlebih dahulu.
|
| 228 |
</p>
|
| 229 |
</div>
|
| 230 |
</div>"""
|
| 231 |
return None, "", warn_html
|
| 232 |
|
| 233 |
-
# ββ
|
| 234 |
-
|
|
|
|
|
|
|
|
|
|
| 235 |
for page in doc:
|
| 236 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 237 |
|
|
|
|
| 238 |
ner_results = ner(full_text.strip())
|
| 239 |
|
| 240 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 241 |
entity_map: dict[str, str] = {}
|
| 242 |
for ent in ner_results:
|
| 243 |
word = clean_word(ent["word"])
|
| 244 |
if len(word) < 2:
|
| 245 |
continue
|
| 246 |
label_id = get_label_id(ent["entity_group"])
|
| 247 |
-
#
|
| 248 |
-
if word not in entity_map
|
| 249 |
-
entity_map[word] = label_id
|
| 250 |
|
| 251 |
if not entity_map:
|
| 252 |
doc.close()
|
| 253 |
return None, "", "<p style='color:#94a3b8; padding:16px;'>Tidak ada entitas yang ditemukan dalam PDF.</p>"
|
| 254 |
|
| 255 |
-
#
|
| 256 |
sorted_entities = sorted(entity_map.items(), key=lambda x: len(x[0]), reverse=True)
|
| 257 |
|
| 258 |
-
# Track which labels actually appeared
|
| 259 |
found_labels: set[str] = set()
|
| 260 |
|
| 261 |
-
#
|
| 262 |
-
for page in doc:
|
| 263 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 264 |
color = LABEL_COLORS.get(label_id, (0.9, 0.9, 0.9))
|
| 265 |
-
instances = page.search_for(
|
| 266 |
for rect in instances:
|
| 267 |
annot = page.add_highlight_annot(rect)
|
| 268 |
annot.set_colors(stroke=list(color))
|
| 269 |
annot.update()
|
| 270 |
found_labels.add(label_id)
|
| 271 |
|
| 272 |
-
#
|
| 273 |
out_fd, out_path = tempfile.mkstemp(suffix=".pdf")
|
| 274 |
os.close(out_fd)
|
| 275 |
doc.save(out_path)
|
| 276 |
doc.close()
|
| 277 |
|
| 278 |
-
#
|
| 279 |
legend_rows = ""
|
| 280 |
for label_id in sorted(found_labels):
|
| 281 |
label_idn = LABEL_MAP.get(label_id, label_id)
|
|
@@ -318,8 +346,7 @@ def run_ner_pdf(pdf_file):
|
|
| 318 |
</div>"""
|
| 319 |
|
| 320 |
return out_path, legend_html, ""
|
| 321 |
-
|
| 322 |
-
|
| 323 |
# Custom CSS
|
| 324 |
CUSTOM_CSS = """
|
| 325 |
@import url('https://fonts.googleapis.com/css2?family=Playfair+Display:wght@700;900&family=DM+Sans:wght@400;500;600&display=swap');
|
|
|
|
| 143 |
|
| 144 |
rows_html = ""
|
| 145 |
row_num = 1
|
| 146 |
+
seen_words = set() # β tambah ini
|
| 147 |
+
|
| 148 |
for ent in results:
|
| 149 |
raw_label = ent["entity_group"]
|
| 150 |
label_id = get_label_id(raw_label)
|
|
|
|
| 153 |
if not word:
|
| 154 |
continue
|
| 155 |
|
| 156 |
+
# ββ Deduplikasi: skip kalau kata ini sudah pernah muncul ββ
|
| 157 |
+
word_key = word.lower()
|
| 158 |
+
if word_key in seen_words:
|
| 159 |
+
continue
|
| 160 |
+
seen_words.add(word_key)
|
| 161 |
+
|
| 162 |
hex_color = LABEL_HEX.get(label_id, "#e2e8f0")
|
| 163 |
score = f"{ent['score']:.2%}"
|
| 164 |
even = row_num % 2 == 0
|
|
|
|
| 179 |
</tr>"""
|
| 180 |
row_num += 1
|
| 181 |
|
| 182 |
+
if not rows_html:
|
| 183 |
+
return "<p style='color:#94a3b8; font-style:italic; padding:16px;'>Tidak ada entitas yang ditemukan.</p>"
|
| 184 |
+
|
| 185 |
table_html = f"""
|
| 186 |
<div style="overflow-x:auto; margin-top:4px; border-radius:12px;
|
| 187 |
border:1px solid #e2e8f0; box-shadow:0 2px 12px rgba(0,0,0,0.06);">
|
|
|
|
| 204 |
</div>"""
|
| 205 |
return table_html
|
| 206 |
|
|
|
|
| 207 |
# PDF NER
|
| 208 |
def run_ner_pdf(pdf_file):
|
|
|
|
|
|
|
|
|
|
| 209 |
if pdf_file is None:
|
| 210 |
return (
|
| 211 |
None,
|
|
|
|
| 214 |
)
|
| 215 |
|
| 216 |
pdf_path = pdf_file if isinstance(pdf_file, str) else pdf_file.name
|
|
|
|
| 217 |
doc = fitz.open(pdf_path)
|
| 218 |
page_count = len(doc)
|
| 219 |
|
|
|
|
| 220 |
if page_count > MAX_PDF_PAGES:
|
| 221 |
doc.close()
|
| 222 |
warn_html = f"""
|
|
|
|
| 229 |
<p style="margin:4px 0 0; color:#9f1239; font-size:13px;">
|
| 230 |
File Anda memiliki <strong>{page_count} halaman</strong>.
|
| 231 |
Maksimal yang diizinkan adalah <strong>{MAX_PDF_PAGES} halaman</strong>.
|
|
|
|
| 232 |
</p>
|
| 233 |
</div>
|
| 234 |
</div>"""
|
| 235 |
return None, "", warn_html
|
| 236 |
|
| 237 |
+
# ββ LANGKAH 1: Kumpulkan semua teks per halaman (simpan, jangan hapus) ββ
|
| 238 |
+
# get_text("words") β list of (x0,y0,x1,y1, word, block_no, line_no, word_no)
|
| 239 |
+
pages_words = [] # per halaman: list of (rect, word_lower)
|
| 240 |
+
full_text_parts = []
|
| 241 |
+
|
| 242 |
for page in doc:
|
| 243 |
+
words_on_page = page.get_text("words")
|
| 244 |
+
pages_words.append(words_on_page)
|
| 245 |
+
full_text_parts.append(page.get_text())
|
| 246 |
+
|
| 247 |
+
full_text = " ".join(full_text_parts)
|
| 248 |
|
| 249 |
+
# ββ LANGKAH 2: Masukkan semua teks ke model (sekali jalan) ββ
|
| 250 |
ner_results = ner(full_text.strip())
|
| 251 |
|
| 252 |
+
if not ner_results:
|
| 253 |
+
doc.close()
|
| 254 |
+
return None, "", "<p style='color:#94a3b8; padding:16px;'>Tidak ada entitas yang ditemukan dalam PDF.</p>"
|
| 255 |
+
|
| 256 |
+
# ββ LANGKAH 3: Bangun entity_map dari output model ββ
|
| 257 |
+
# key: kata/frasa (lowercase), value: label_id
|
| 258 |
entity_map: dict[str, str] = {}
|
| 259 |
for ent in ner_results:
|
| 260 |
word = clean_word(ent["word"])
|
| 261 |
if len(word) < 2:
|
| 262 |
continue
|
| 263 |
label_id = get_label_id(ent["entity_group"])
|
| 264 |
+
# Kalau sudah ada, prioritaskan frasa yang lebih panjang
|
| 265 |
+
if word.lower() not in entity_map:
|
| 266 |
+
entity_map[word.lower()] = label_id
|
| 267 |
|
| 268 |
if not entity_map:
|
| 269 |
doc.close()
|
| 270 |
return None, "", "<p style='color:#94a3b8; padding:16px;'>Tidak ada entitas yang ditemukan dalam PDF.</p>"
|
| 271 |
|
| 272 |
+
# Urutkan dari yang terpanjang agar frasa multi-kata dicocokan duluan
|
| 273 |
sorted_entities = sorted(entity_map.items(), key=lambda x: len(x[0]), reverse=True)
|
| 274 |
|
|
|
|
| 275 |
found_labels: set[str] = set()
|
| 276 |
|
| 277 |
+
# ββ LANGKAH 4 & 5: Cocokan + warnai β per halaman, bukan per entitas ββ
|
| 278 |
+
for page_idx, page in enumerate(doc):
|
| 279 |
+
words_on_page = pages_words[page_idx]
|
| 280 |
+
|
| 281 |
+
# Bangun teks halaman lowercase untuk window-matching
|
| 282 |
+
# Kita pakai search_for PyMuPDF tapi HANYA untuk entitas yang
|
| 283 |
+
# memang ADA di halaman ini β cek dulu dengan `in page_text`
|
| 284 |
+
page_text_lower = doc[page_idx].get_text().lower()
|
| 285 |
+
|
| 286 |
+
for entity_lower, label_id in sorted_entities:
|
| 287 |
+
# ββ Langkah 4: cek keberadaan dulu (O(n) string search, murah) ββ
|
| 288 |
+
if entity_lower not in page_text_lower:
|
| 289 |
+
continue # skip, tidak ada di halaman ini
|
| 290 |
+
|
| 291 |
+
# ββ Langkah 5: baru panggil search_for (lebih mahal) ββ
|
| 292 |
color = LABEL_COLORS.get(label_id, (0.9, 0.9, 0.9))
|
| 293 |
+
instances = page.search_for(entity_lower, quads=False)
|
| 294 |
for rect in instances:
|
| 295 |
annot = page.add_highlight_annot(rect)
|
| 296 |
annot.set_colors(stroke=list(color))
|
| 297 |
annot.update()
|
| 298 |
found_labels.add(label_id)
|
| 299 |
|
| 300 |
+
# Simpan ke file temp
|
| 301 |
out_fd, out_path = tempfile.mkstemp(suffix=".pdf")
|
| 302 |
os.close(out_fd)
|
| 303 |
doc.save(out_path)
|
| 304 |
doc.close()
|
| 305 |
|
| 306 |
+
# ββ Legend ββ
|
| 307 |
legend_rows = ""
|
| 308 |
for label_id in sorted(found_labels):
|
| 309 |
label_idn = LABEL_MAP.get(label_id, label_id)
|
|
|
|
| 346 |
</div>"""
|
| 347 |
|
| 348 |
return out_path, legend_html, ""
|
| 349 |
+
|
|
|
|
| 350 |
# Custom CSS
|
| 351 |
CUSTOM_CSS = """
|
| 352 |
@import url('https://fonts.googleapis.com/css2?family=Playfair+Display:wght@700;900&family=DM+Sans:wght@400;500;600&display=swap');
|