Update app.py

app.py

@@ -1,18 +1,39 @@
 
 import gradio as gr
 
-
 import numpy as np
 import pandas as pd
-
 from sklearn.model_selection import train_test_split
-from simpletransformers.classification import ClassificationModel
+
+import tensorflow as tf
+from transformers import BertTokenizer, TFBertModel
+from tensorflow.keras.layers import Dense
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.models import load_model
+
 from sklearn.metrics import classification_report,confusion_matrix
 import re
 import nltk
 from nltk.corpus import stopwords
 nltk.download('stopwords')
 
+# Load the tokenizer and model
+tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
+bert_model = TFBertModel.from_pretrained("bert-base-uncased")
+
+# Define function to create embeddings
+def bert_embeddings(texts, max_length=128):
+    inputs = tokenizer(
+        texts.tolist(),
+        return_tensors="tf",
+        padding=True,
+        truncation=True,
+        max_length=max_length
+    )
+    outputs = bert_model(inputs['input_ids'], attention_mask=inputs['attention_mask'])
+    cls_embeddings = outputs.last_hidden_state[:, 0, :]  # CLS token's embedding
+    return cls_embeddings
+
 file_path = "https://raw.githubusercontent.com/alexvatti/full-stack-data-science/main/NLP-Exercises/Movie-Review/IMDB%20Dataset.csv"
 movies_df=pd.read_csv(file_path)
 
@@ -37,46 +58,46 @@ movies_df['review'] = movies_df['review'].apply(remove_stop_wrods)
 movies_df["Category"]=movies_df["sentiment"].apply(lambda x: 1 if x=='positive' else 0)
 
 X_train,X_test,y_train,y_test=train_test_split(movies_df['review'],movies_df["Category"],test_size=0.2,random_state=42)
-# Prepare the training and evaluation DataFrames for Simple Transformers
-train_df = pd.DataFrame({"text": X_train, "labels": y_train})
-eval_df = pd.DataFrame({"text": X_test, "labels": y_test})
-
+# Convert emails to BERT embeddings
+X_train_embeddings = bert_embeddings(X_train)
+X_test_embeddings = bert_embeddings(X_test)
 
-# Create a ClassificationModel
-model = ClassificationModel("bert", "bert-base-uncased", use_cuda=False)  # Set use_cuda=True if you have a GPU
+# Define a simple classifier model
+classifier = Sequential([
+    Dense(128, activation='relu', input_shape=(768,)),
+    Dense(1, activation='sigmoid')  # Sigmoid for binary classification
+])
 
-# Train the model
-model.train_model(train_df)
+# Compile the classifier
+classifier.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
 
-# Evaluate the model
-result, model_outputs, wrong_predictions = model.eval_model(eval_df)
-model.save_model("sentiment_model")
+# Train the classifier
+classifier.fit(X_train_embeddings, y_train, epochs=5, batch_size=32, validation_split=0.1)
 
-# Step 4: Load the Model for Prediction
-# To use the model later, reload it from the saved directory
-loaded_model = ClassificationModel("bert", "sentiment_model", use_cuda=True)
+# Evaluate on test set
+test_loss, test_accuracy = classifier.evaluate(X_test_embeddings, y_test)
+print(f"Test Accuracy: {test_accuracy}")
 
-# Step 5: Predict Sentiment for a New Review
-test_review = "This movie was absolutely fantastic! The acting was top-notch."
-review=remove_tags(test_review)
-review=remove_stop_wrods(review)
-predictions, raw_outputs = loaded_model.predict(review)
-print("Predictions:", predictions)  # Outputs the label (e.g., 1 for positive, 0 for negative)
-print("Raw Outputs:", raw_outputs)  # Outputs the raw model scores
+# Predictions and confusion matrix
+y_pred = (classifier.predict(X_test_embeddings) > 0.5).astype("int32")
+conf_matrix = confusion_matrix(y_test, y_pred)
+class_report = classification_report(y_test, y_pred)
 
-test_review ="I hated this movie. It was a complete waste of time."
-review=remove_tags(test_review)
-review=remove_stop_wrods(review)
-predictions, raw_outputs = loaded_model.predict(review)
+print("Confusion Matrix:")
+print(conf_matrix)
+print("\nClassification Report:")
+print(class_report)
 
-print("Predictions:", predictions)  # Outputs the label (e.g., 1 for positive, 0 for negative)
-print("Raw Outputs:", raw_outputs)  # Outputs the raw model scores
+# Save the trained model to a file
+classifier.save("movie_sentiment_model.h5")
 
 def fn(test_review):
     review=remove_tags(test_review)
     review=remove_stop_wrods(review)
-    predictions, raw_outputs = loaded_model.predict(review)
-    return "Positive" if predictions==1 else "Negative"
+    cls_embeddings = bert_embeddings([review])
+    loaded_model = load_model("movie_sentiment_model.h5")
+    prediction = loaded_model.predict(cls_embeddings)
+    return "Positive" if prediction[0] > 0.5 else "Negative"
     
 description = "Give a review of a movie that you like(or hate, sarcasm intended XD) and the model will let you know just how much your review truely reflects your emotions. "
 input_text = gr.Textbox(label="Enter Text")