Update helper.py

helper.py

@@ -2,7 +2,6 @@ import demoji
 import re
 import pandas as pd
 import numpy as np
-import nltk
 import keras
 import os
 import warnings
@@ -11,7 +10,6 @@ import tensorflow as tf
 from transformers import BertTokenizer, TFBertModel
 from string import punctuation
 from keybert import KeyBERT
-from nltk.corpus import stopwords
 from sentence_transformers import SentenceTransformer
 
 # --- Disable warnings ---
@@ -22,16 +20,16 @@ MAX_LENGTH = 128
 base_path = os.path.join('data')
 model_path = os.path.join('Model')
 
-# --- Setup NLTK data directory (custom path for Hugging Face) ---
-nltk_data_dir = os.path.join(tempfile.gettempdir(), "nltk_data")
-os.makedirs(nltk_data_dir, exist_ok=True)
-nltk.data.path.append(nltk_data_dir)
-nltk.download('stopwords', download_dir=nltk_data_dir)
-
 # --- Load Resources ---
 alay_dict = pd.read_csv(os.path.join(base_path, 'kamus_alay.csv'), names=['alay', 'normal'], encoding='latin-1')
 alay_dict_map = dict(zip(alay_dict['alay'], alay_dict['normal']))
-stop_words = set(stopwords.words('indonesian'))
+
+# Load stopwords dari Excel
+stopwords_excel_path = os.path.join(base_path, 'stopwords_indonesia.xlsx')
+stopwords_df = pd.read_excel(stopwords_excel_path)
+stop_words = set(stopwords_df['stopword'].astype(str).tolist())
+
+# Load tokenizer & model
 tokenizer = BertTokenizer.from_pretrained("indobenchmark/indobert-large-p1")
 bert_model = TFBertModel.from_pretrained("indobenchmark/indobert-large-p1")
 lstm_model = keras.models.load_model(os.path.join(model_path, 'indobert_lstm_model.keras'))