pada kode analisis sentimen berikut, tambahkan model CNN-LSTM ke kodenya dan atur…
- User
pada kode analisis sentimen berikut, tambahkan model CNN-LSTM ke kodenya dan atur sedemikian rupa, tambahkan kode lain jika diperlukan! Kode analisis sentimen: import pickle import pandas as pd import numpy as np import torch from torch.utils.data import Dataset from sklearn.model_selection import train_test_split from sklearn.metrics import classification_report, accuracy_score
from transformers import AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
Membaca dataset
data_list = pickle.load(open('filtered_pre_processed_berita_joined_done.pkl', 'rb')) data = pd.DataFrame(data_list, columns=['judul', 'isi', 'pre_processed', 'Label', 'Nama_Pelabel']) data['Isi Berita'] = data['pre_processed']
Tentukan label
sentiment_mapping = {'Positif': 0, 'Negatif': 1, 'Netral': 2} data['Label'] = data['Label'].map(sentiment_mapping)
Split dataset
train_texts, test_texts, train_labels, test_labels = train_test_split(data['Isi Berita'].tolist(), data['Label'].tolist(), test_size=0.20, random_state=42)
Load tokenizer dan model
tokenizer = AutoTokenizer.from_pretrained('indolem/indobert-base-uncased') model = AutoModelForSequenceClassification.from_pretrained('indolem/indobert-base-uncased', num_labels=len(sentiment_mapping))
Atur dataset
class NewsDataset(Dataset): def init(self, texts, labels, tokenizer, max_len): self.texts = texts self.labels = labels self.tokenizer = tokenizer self.max_len = max_len
def __len__(self): return len(self.texts) def __getitem__(self, index): text = str(self.texts[index]) label = self.labels[index] inputs = self.tokenizer.encode_plus( text, add_special_tokens=True, max_length=self.max_len, padding='max_length', truncation=True ) return {'input_ids': torch.tensor(inputs['input_ids'], dtype=torch.long), 'attention_mask': torch.tensor(inputs['attention_mask'], dtype=torch.long), 'labels': torch.tensor(label, dtype=torch.long) }max_len = 400 train_dataset = NewsDataset(train_texts, train_labels, tokenizer, max_len) test_dataset = NewsDataset(test_texts, test_labels, tokenizer, max_len)"
Kode CNN-LSTM: " # Membangun model untuk lipatan saat ini model = Sequential([ Embedding(input_dim=10000, output_dim=300, input_length=500, weights=[embedding_matrix], trainable=False), Conv1D(256, 3, activation='relu', kernel_regularizer=l2(0.01)), GlobalMaxPooling1D(), Dropout(0.5), Reshape((-1, 256)), LSTM(200, return_sequences=False, kernel_regularizer=l2(0.01), recurrent_regularizer=l2(0.01)), Dropout(0.7), Dense(32, activation='relu', kernel_regularizer=l2(0.01)), Dropout(0.5), Dense(3, activation='softmax') ])
# Buat callback early stopping callback_es = EarlyStopping(monitor='val_loss', patience=5, verbose=1, restore_best_weights=True) initial_learning_rate = 0.0001 optimizer = Adam(learning_rate=initial_learning_rate) # Kompilasi model model.compile(optimizer=optimizer, loss='categorical_crossentropy', metrics=['accuracy']) # Latih model dengan menggunakan learning rate scheduler dan early stopping dalam callbacks history = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=100, batch_size=128, callbacks=[callback_es])"di akhir tambahkan kode akurasi dan classification report: "# Menguji model predictions, true_labels, _ = trainer.predict(test_dataset) predicted_labels = np.argmax(predictions, axis=1) print(classification_report(test_labels, predicted_labels)) print("Accuracy: ", accuracy_score(test_labels, predicted_labels))"
- Assistant
Untuk menambahkan model CNN-LSTM ke kode analisis sentimen di atas, kita perlu membuat beberapa modifikasi pada kode. Pertama, kita bisa menggunakan
Tokenizerdari transformer untuk menggantikan metode preprocessing yang ada. Kemudian, kita perlu menggantikan modelSequentialdengan model LSTM dan CNN baru. Berikut adalah kode yang telah dimodifikasi:import pickle import pandas as pd import numpy as np import torch from torch.utils.data import Dataset from sklearn.model_selection import train_test_split from sklearn.metrics import classification_report, accuracy_score from transformers import AutoTokenizer, TrainingArguments, Trainer from keras.utils import to_categorical from keras.preprocessing.sequence import pad_sequences from keras.models import Sequential from keras.layers import Embedding, Conv1D, GlobalMaxPooling1D, Dropout, Reshape, LSTM, Dense from keras.callbacks import EarlyStopping from keras.optimizers import Adam from keras.regularizers import l2 # Membaca dataset data_list = pickle.load(open('filtered_pre_processed_berita_joined_done.pkl', 'rb')) data = pd.DataFrame(data_list, columns=['judul', 'isi', 'pre_processed', 'Label', 'Nama_Pelabel']) data['Isi Berita'] = data['pre_processed'] # Tentukan label sentiment_mapping = {'Positif': 0, 'Negatif': 1, 'Netral': 2} data['Label'] = data['Label'].map(sentiment_mapping) # Split dataset train_texts, test_texts, train_labels, test_labels = train_test_split(data['Isi Berita'].tolist(), data['Label'].tolist(), test_size=0.20, random_state=42) # Load tokenizer dan model tokenizer = AutoTokenizer.from_pretrained('indolem/indobert-base-uncased') def encode(texts, tokenizer, max_len=500): input_ids = [] for text in texts: input_dict = tokenizer.encode_plus(text, add_special_tokens=True, max_length=max_len, padding='max_length', truncation=True) input_ids.append(input_dict['input_ids']) return np.array(input_ids) train_input_ids = encode(train_texts, tokenizer) test_input_ids = encode(test_texts, tokenizer) # One-hot encode labels train_one_hot_labels = to_categorical(train_labels) test_one_hot_labels = to_categorical(test_labels) # CNN-LSTM model model = Sequential([ Embedding(input_dim=tokenizer.vocab_size, output_dim=300, input_length=500, mask_zero=True, ), Conv1D(256, 3, activation='relu', kernel_regularizer=l2(0.01)), GlobalMaxPooling1D(), Dropout(0.5), Reshape((-1, 256)), LSTM(200, return_sequences=False, kernel_regularizer=l2(0.01), recurrent_regularizer=l2(0.01)), Dropout(0.7), Dense(32, activation='relu', kernel_regularizer=l2(0.01)), Dropout(0.5), Dense(len(sentiment_mapping), activation='softmax') ]) callback_es = EarlyStopping(monitor='val_loss', patience=5, verbose=1, restore_best_weights=True) initial_learning_rate = 0.0001 optimizer = Adam(learning_rate=initial_learning_rate) # Kompilasi model model.compile(optimizer=optimizer, loss='categorical_crossentropy', metrics=['accuracy']) # Latih model history = model.fit(train_input_ids, train_one_hot_labels, validation_data=(test_input_ids, test_one_hot_labels), epochs=100, batch_size=128, callbacks=[callback_es]) # Menguji model predictions = model.predict(test_input_ids) predicted_labels = np.argmax(predictions, axis=1) print(classification_report(test_labels, predicted_labels)) print("Accuracy: ", accuracy_score(test_labels, predicted_labels))Perlu dicatat bahwa dalam contoh ini, kita menggunakan tokenizer dari model IndoBERT dan tidak menggunakan pretrained embedding. Anda dapat menyesuaikan kode ini sesuai kebutuhan Anda, misalnya, dengan memasukkan pretrained embedding atau mengubah parameter lainnya.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.