Update app.py

added logic to classify the user input as one of the five faq response keys, else generate a response from the chatbot model

app.py

@@ -4,6 +4,9 @@ import gradio as gr  # Import Gradio for the interface
 # Load a text-generation model
 chatbot = pipeline("text-generation", model="microsoft/DialoGPT-medium")
 
+# Load the classification model
+classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
+
 # Customize the bot's knowledge base with predefined responses
 faq_responses = {
     "study tips": "Here are some study tips: 1) Break your study sessions into 25-minute chunks (Pomodoro Technique). 2) Test yourself frequently. 3) Stay organized using planners or apps like Notion or Todoist.",
@@ -15,10 +18,25 @@ faq_responses = {
 
 # Define the chatbot's response function
 def faq_chatbot(user_input):
+    # Classify the user input by passing the FAQ keywords as labels
+    classified_user_input = classifier(user_input, candidate_labels=list(faq_responses.keys()))
+
+    # Get the highest confidence score label, ie. the most likely of the FAQ
+    predicted_label = classified_user_input["labels"][0]
+    confidence_score = classified_user_input["scores"][0]
+
+    # Confidence threshold (adjust if needed)
+    threshold = 0.5
+
+    # If the classification confidence is high, return the corresponding FAQ response
+    if confidence_score > threshold:
+        return faq_responses[predicted_label]
+    
+    
     # Check if the user's input matches any FAQ keywords
-    for key, response in faq_responses.items():
-        if key in user_input.lower():
-            return response
+    # for key, response in faq_responses.items():
+    #     if key in user_input.lower():
+    #         return response
 
     # If no FAQ match, use the AI model to generate a response
     conversation = chatbot(user_input, max_length=50, num_return_sequences=1)