Creating Smart Chatbots Using NLP: A Complete Beginner’s Guide to Intelligent Conversational Agents

📗 Chapter 5: Adding Memory, Personalization, and Context Awareness

Build Chatbots That Remember, Adapt, and Truly Understand

🧠 Introduction

A great chatbot doesn’t just answer questions — it remembers who you are, adapts to your preferences, and continues the conversation as if it never ended. This capability comes from memory, personalization, and context awareness.

The future of conversational AI lies in bots that understand the history and individuality of each user.

In this chapter, you’ll learn how to:

• Track memory and store user data
• Use context to manage flow and tone
• Personalize responses with preferences
• Implement these features using Rasa, Python, or custom frameworks
• Secure and ethically manage user data

Let’s take your bot from “useful” to “unforgettable.”

📘 Section 1: What Is Context Awareness and Memory?

📘 Section 2: Why Do Chatbots Need Memory?

Without memory:

• Bots feel repetitive
• Every conversation feels like “starting over”
• They fail in multi-turn interactions

With memory:

• Bots track name, intent, location, or purchase history
• They resume broken conversations
• They offer consistent and human-like UX

📘 Section 3: Tracking User State with Python

A simple memory tracker using Python:

python

user_memory = {}

def handle_message(user_id, message):

    if user_id not in user_memory:

        user_memory[user_id] = {}

    if "name is" in message.lower():

        name = message.split("name is")[-1].strip()

        user_memory[user_id]["name"] = name

        return f"Nice to meet you, {name}!"

    if "what's my name" in message.lower():

        return f"You're {user_memory[user_id].get('name', 'not introduced yet')}."

    return "I'm here to help you."

📘 Section 4: Using Rasa Slots for Context Memory

🔹 domain.yml

yaml

slots:

  name:

    type: text

    influence_conversation: true

🔹 nlu.yml

yaml

- intent: inform_name

  examples: |

    - My name is [John](name)

    - Call me [Alice](name)

🔹 actions.py

python

class ActionRememberName(Action):

    def name(self):

        return "action_remember_name"

    def run(self, dispatcher, tracker, domain):

        name = tracker.get_slot("name")

        dispatcher.utter_message(f"Hello again, {name}!")

        return []

📘 Section 5: Multi-Turn Context Tracking

Let’s track context like booking, canceling, and asking for help — without repeating earlier questions.

✅ Rasa: Context-Aware Conversation Example

yaml

- story: personalized greeting

  steps:

    - intent: greet

    - action: utter_greet

    - intent: inform_name

    - action: action_remember_name

    - intent: ask_flight_status

    - action: action_flight_status

Bot remembers name across intents, even after interruptions.

📘 Section 6: Personalization Based on Preferences

Sample personalization table:

Example response:

python

dispatcher.utter_message(

    f"Welcome back, {tracker.get_slot('name')}! Ready to book another trip to {tracker.get_slot('user_city')}?"

)

📘 Section 7: Using Session Storage and APIs for Memory

Memory Types:

Store user preferences in Redis:

python

import redis

r = redis.StrictRedis(host='localhost', port=6379, decode_responses=True)

def save_user_preference(user_id, key, value):

    r.hset(user_id, key, value)

def get_user_preference(user_id, key):

    return r.hget(user_id, key)

📘 Section 8: Handling Context Switches

A context-aware bot should:

• Detect when the user changes topic
• Pause current task and resume later
• Store progress of the interrupted flow

Example:

“Book me a flight… actually, what’s the weather?”

Store flight intent as pending, answer weather, then resume flight booking.

📘 Section 9: Ethical Memory Use

📌 Important Considerations:

• Consent: Inform users their data is stored
• Clear opt-outs: “Forget me” commands
• Data security: Use encrypted storage, avoid PII unless necessary
• Minimalism: Store only what’s needed for value delivery

📘 Section 10: Real-World Use Case: Food Delivery Chatbot

Features:

• Remembers previous order
• Stores delivery address and payment method
• Suggests similar dishes
• Greets user with name
• Recovers abandoned carts

Tech Stack:

• Python + Rasa
• Redis or SQLite for memory
• Webhook for live updates
• Personalized push notifications

✅ Chapter Summary Table