Update app.py

app.py

@@ -2,6 +2,7 @@ import streamlit as st
 import spacy
 import torch
 import torch.nn as nn
+import pandas as pd
 from transformers import BertTokenizer, BertModel, AutoConfig
 from transformers.models.bert.modeling_bert import BertForMaskedLM
 
@@ -213,11 +214,59 @@ def load_reviews_from_file(file_path):
         st.error(f"File not found: {file_path}")
     return reviews
 
+# Function to load reviews from a CSV file
+def load_reviews_from_csv(file_path):
+    try:
+        df = pd.read_csv(file_path)
+        return df
+    except FileNotFoundError:
+        st.error(f"File not found: {file_path}")
+        return None
+
+# Function to process each review in the CSV and get the model's predictions
+def process_csv_reviews(df):
+    true_reviews = []
+    for _, row in df.iterrows():
+        review_text = row['review']
+        label = row['label']
+        
+        # Get BERT embedding for the review text
+        bert_embedding = get_bert_embedding(review_text.lower())
+        
+        # Get SpaBERT embedding for geo-entities
+        spaBert_embedding, _ = processSpatialEntities(review_text, nlp)
+        
+        # Concatenate BERT and SpaBERT embeddings
+        combined_embedding = torch.cat((bert_embedding, spaBert_embedding), dim=-1)
+        
+        # Get model prediction
+        prediction = get_prediction(combined_embedding)
+        
+        # If prediction is "Not Spam" (0), store the review
+        if prediction == 0:
+            true_reviews.append((review_text, label))
+    
+    # Convert to a DataFrame for easy display
+    return pd.DataFrame(true_reviews, columns=['Review', 'Label'])
 
+st.write("### Process Filtered Reviews CSV")
+csv_file_path = "models/spabert/datasets/filtered_reviews.csv"
 
-
-
-
+if st.button("Process CSV and Find True Reviews"):
+    # Load the CSV file
+    df = load_reviews_from_csv(csv_file_path)
+    
+    if df is not None:
+        # Filter reviews predicted to be "Not Spam"
+        true_reviews_df = process_csv_reviews(df)
+        
+        if not true_reviews_df.empty:
+            st.write("### Reviews Predicted to be Not Spam:")
+            st.dataframe(true_reviews_df)
+        else:
+            st.write("No reviews were predicted to be Not Spam.")
+    else:
+        st.error("Could not load CSV file.")
 
 
 #Demo Section