Abstract

This document presents a chatbot implementation using NLP and ML. The chatbot predicts user

intents and generates

appropriate responses. Features include synonym-based sentence augmentation using WordNet,

intent classification

with a TF-IDF vectorizer and Naive Bayes classifier, and interactive real-time usage.

Explanation of the Source Code

1. Libraries:

- NLTK for lexical resources and synonym generation.

- Scikit-learn for TF-IDF vectorization and Naive Bayes classification.

2. Predefined intents include greetings, weather, and more.

3. Training data is expanded using synonyms to improve classification.

4. A TF-IDF vectorizer converts text into numerical data, which is classified using a Naive Bayes

model.

5. A command-line interface enables real-time interaction with the chatbot.

Source Code

import random

import nltk

from nltk.corpus import wordnet as wn

from sklearn.feature_extraction.text import TfidfVectorizer

from sklearn.pipeline import make_pipeline

import pickle

# Download required NLTK data

nltk.download('wordnet')

nltk.download('omw-1.4')

# Predefined Intents and Responses

intents = {

"greeting": ["hi", "hello", "hey", "good morning", "good evening", "what's up", "howdy", "hi there",

"hello there"],

"ask_bot_status": ["how are you", "how's it going", "what?s up", "how are you doing", "how?s life",

"how?s everything"],

"ask_name": ["what?s your name", "who are you", "what do you go by", "can you tell me your

name", "are you named anything"],

"farewell": ["bye", "goodbye", "see you later", "take care", "good night", "catch you later"],

"love": ["i love you", "i love you 3000", "i care about you", "you are amazing", "you're awesome"],

"marriage": ["will you marry me", "will you be my life partner", "let's get married", "will you marry

me, chatbot"],

"joke": ["tell me a joke", "make me laugh", "tell me something funny", "crack me up", "give me a

funny line"],

"help": ["help", "can you assist me", "what can you do", "can you help me with something", "I need

your help"],

"weather": ["what?s the weather like today", "is it sunny outside", "how?s the weather", "will it rain

today", "what's the temperature"],

"time": ["what time is it", "what?s the time", "do you know the time", "can you tell me the time"]
}

# Function to generate synonyms using WordNet

def get_synonyms(word):

synonyms = set()

for syn in wn.synsets(word):

for lemma in syn.lemmas():

synonyms.add(lemma.name())

return list(synonyms)

# Function to create multiple variations of a sentence

def generate_variations(sentence):

words = sentence.split()

variations = []

for i, word in enumerate(words):

# Get synonyms for each word

synonyms = get_synonyms(word)

if synonyms:

# Replace the word with a random synonym

new_word = random.choice(synonyms)

new_sentence = words[:i] + [new_word] + words[i+1:]

variations.append(" ".join(new_sentence))

return variations

# Function to expand training data

training_data = []

# Iterate over each intent and sentence

for intent, sentences in intents.items():

for sentence in sentences:

# Add original sentence

training_data.append({"intent": intent, "text": sentence})

# Generate and add variations

variations = generate_variations(sentence)

for variation in variations:

training_data.append({"intent": intent, "text": variation})

return training_data

# Prepare the training data and labels

def prepare_data(training_data):

texts = [item["text"] for item in training_data]

labels = [item["intent"] for item in training_data]

return texts, labels

# Train the machine learning model

def train_model(texts, labels):

model = make_pipeline(TfidfVectorizer(), MultinomialNB())

model.fit(texts, labels)

return model
# Function to predict the intent of the user's message

def predict_intent(user_input, model):

prediction = model.predict([user_input])

return prediction[0]

# Responses for each intent

responses = {

"greeting": ["Hello!", "Hi there!", "Hey!", "Greetings!", "Good to see you!", "Howdy!"],

"ask_bot_status": ["I'm doing great, thanks for asking!", "I'm just a bot, but I'm doing fine.", "I'm

feeling great!"],

"ask_name": ["I'm a chatbot!", "I don't have a name, but you can call me Bot.", "I'm your friendly

assistant!"],

"help": ["I can help you with a lot of things!", "Ask me anything, and I'll try my best to assist you.",

"How can I assist you today?"],

"farewell": ["Goodbye!", "See you later!", "Take care!", "Catch you later!", "Bye for now!"],

"love": ["I love you too", "Love you too!", "I?m flattered, thank you!"],

"marriage": ["Sorry, I can't marry you. I'm just a chatbot.", "I can't marry you, but I can be a good

virtual friend."],

"joke": ["Why don't skeletons fight each other? They don't have the guts!", "Why did the chicken

cross the road? To get to the other side!"],

"weather": ["I can't fetch real-time weather, but you can check a weather website or app for the

latest forecast!"],

"time": ["Sorry, I can't tell you the exact time, but it?s always a good time to chat!"]

# Function to get a response based on the predicted intent

return random.choice(responses.get(intent, ["Sorry, I didn't understand that."]))

# Main function for the chatbot

def chatbot(model):

print("Hello! I'm your chatbot. Type 'exit' to quit.")

while True:

# Get input from the user

user_input = input("You: ")

# Exit condition

if user_input.lower() == "exit":

print("Goodbye!")

break

# Predict the intent of the user input

intent = predict_intent(user_input, model)

# Get a response based on the predicted intent

response = get_response(intent)

# Print the response

print("Bot: " + response)

# Generate the training data

training_data = expand_training_data()
# Prepare the data for training

texts, labels = prepare_data(training_data)

# Train the model

model = train_model(texts, labels)

# Now you can start the chatbot with the trained model

chatbot(model)