Chatbot Product Recommendations using NLP

 Chatbot Product Recommendations using NLP

Business Objective: Provide instant assistance to customers on product recommendations on a Customer Service Chatbot at eCommerce business

Tip: Try it in Google Colab!

1. Generate Dataset

import pandas as pd

# Create a synthetic dataset for customer queries and interactions

data = {

    'query': ['How can I track my order?', 'Do you have this item in stock?', 'Can I return this product?',

              'Which products are trending?', 'Can you recommend a laptop for programming?'],

    'intent': ['order_tracking', 'product_availability', 'return_policy',

               'trending_products', 'product_recommendation'],

    'response': ['To track your order, please visit our website and go to the order tracking section.',

                 'Yes, the item is currently in stock.',

                 'Our return policy allows returns within 30 days of purchase.',

                 'Our trending products include electronics, clothing, and accessories.',

                 'Sure! Based on your needs, we recommend the XYZ laptop for programming.']

}

# Create a DataFrame from the data

customer_data = pd.DataFrame(data)

# Display the dataset

print("Customer Service Dataset:")

print(customer_data)

Model Building- NLP

from sklearn.feature_extraction.text import TfidfVectorizer

from sklearn.metrics.pairwise import cosine_similarity

from sklearn.pipeline import Pipeline

from sklearn.naive_bayes import MultinomialNB

# Define a TF-IDF Vectorizer

tfidf_vectorizer = TfidfVectorizer()

# Define a Naive Bayes Classifier

naive_bayes_classifier = MultinomialNB()

# Create an NLP Pipeline

nlp_pipeline = Pipeline([

    ('tfidf_vectorizer', tfidf_vectorizer),

    ('classifier', naive_bayes_classifier)

])

# Fit the pipeline on the dataset

nlp_pipeline.fit(customer_data['query'], customer_data['intent'])

# Function to predict intent based on query

def predict_intent(query):

    return nlp_pipeline.predict([query])[0]

# Test the model

query = "Can you recommend a laptop for gaming?"

predicted_intent = predict_intent(query)

print(f"Predicted Intent for query '{query}': {predicted_intent}")

Show and Save the Results

# Predict intents for a list of queries

queries = ["How do I track my order?", "Is this item available for shipping?", "What is your refund policy?"]

predicted_intents = [predict_intent(q) for q in queries]

# Display the results

print("\nPredicted Intents for Queries:")

for i, query in enumerate(queries):

    print(f"Query: {query} | Predicted Intent: {predicted_intents[i]}")

# Save the results to a CSV file

results_df = pd.DataFrame({'Query': queries, 'Predicted Intent': predicted_intents})

results_df.to_csv('nlp_results.csv', index=False)

print("\nResults saved to 'nlp_results.csv' file.")

Get the code: Github