Deploy a Chatbot using TensorFlow in Python

Last Updated : 23 Jul, 2025

In this article, you'll learn how to deploy a Chatbot using Tensorflow. A Chatbot is basically a bot (a program) that talks and responds to various questions just like a human would. We'll be using a number of Python modules to do this.

This article is divided into two sections:

First, we'll train the Chatbot model, and then in section two, we'll learn how to make it work and respond to various inputs by the user.

Modules required:

random - This module is used to generate random responses from the Chatbot
json - To read from json file
pickle - To save data into files
tensorflow - To train neural networks. It is an open source machine learning library.
numpy - It is a Python library used for working with arrays
nltk - It is a leading platform for building Python programs to work with human language data.

Training the Chatbot model

The first thing we're going to do is to train the Chatbot model. In order to do that, create a file named 'intense.json' in which we'll write all the intents, tags and words or phrases our Chatbot would be responding to.

Step 1: Create a "intense.json":

Here we've added just three tags just to show how it works. You can add a lot of them!

Step 2: Create a "training.py":

The next step is training our model. We'll use a class called WordNetLemmatizer() which will give the root words of the words that the Chatbot can recognize. For example, for hunting, hunter, hunts and hunted, the lemmatize function of the WordNetLemmatizer() class will give "hunt" because it is the root word.

Create a WordNetLemmatizer() class object.
Read the contents from the "intense.json" file and store it to a variable "intents". Next, initialize empty lists to store the contents.
Next up, we have a function called word_tokenize(para). It takes a sentence as a parameter and then returns a list containing all the words of the sentence as strings. Here we're tokenizing the patterns and then appending them to a list 'words'. So, at last, this list 'words' would have all the words that are in the 'patterns' list.
In documents, we have all the patterns with their tags in the form of a tuple.
Now, using a list comprehension, we'll modify the list 'words' we created above and store the words' 'lemma' or simply put, the root words.
Dump the data of the 'words' and 'classes' to binary files of the same name, using the pickle module's dump() function.

Python3

# importing the required modules.
import random
import json
import pickle
import numpy as np
import nltk

from keras.models import Sequential
from nltk.stem import WordNetLemmatizer
from keras.layers import Dense, Activation, Dropout
from keras.optimizers import SGD


lemmatizer = WordNetLemmatizer()

# reading the json.intense file
intents = json.loads(open("intense.json").read())

# creating empty lists to store data
words = []
classes = []
documents = []
ignore_letters = ["?", "!", ".", ","]
for intent in intents['intents']:
    for pattern in intent['patterns']:
        # separating words from patterns
        word_list = nltk.word_tokenize(pattern)
        words.extend(word_list)  # and adding them to words list
        
        # associating patterns with respective tags
        documents.append(((word_list), intent['tag']))

        # appending the tags to the class list
        if intent['tag'] not in classes:
            classes.append(intent['tag'])

# storing the root words or lemma
words = [lemmatizer.lemmatize(word)
         for word in words if word not in ignore_letters]
words = sorted(set(words))

# saving the words and classes list to binary files
pickle.dump(words, open('words.pkl', 'wb'))
pickle.dump(classes, open('classes.pkl', 'wb'))

Step 3: Now we need to classify our data into 0's and 1's because neural networks works with numerical values, not strings or anything else.

Create an empty list called training, in which we'll store the data used for training. Also create an output_empty list that will store as many 0's as there are classes in the intense.json.
Next up we'll create a bag that will store the 0's and 1's. (0, if the word isn't in the pattern and 1 if the word is in the pattern). To do that, we'll iterate through the documents list and append 1 to the 'bag' if it is not in the patterns, 0 otherwise.
Now shuffle this training set and make it a numpy array.
Split the training set consisting of 1's and 0's into two parts, that is train_x and train_y.

Python3

# we need numerical values of the
# words because a neural network
# needs numerical values to work with
training = []
output_empty = [0]*len(classes)
for document in documents:
    bag = []
    word_patterns = document[0]
    word_patterns = [lemmatizer.lemmatize(
        word.lower()) for word in word_patterns]
    for word in words:
        bag.append(1) if word in word_patterns else bag.append(0)
        
    # making a copy of the output_empty
    output_row = list(output_empty)
    output_row[classes.index(document[1])] = 1
    training.append([bag, output_row])
random.shuffle(training)
training = np.array(training)

# splitting the data
train_x = list(training[:, 0])
train_y = list(training[:, 1])

Step 4: We've come to the model-building part of our Chatbot model. Here, we're going to deploy a Sequential model, that we'll train on the dataset we prepared above.

Add(): This function is used to add layers in a neural network.
Dropout(): This function is used to avoid overfitting

Python3

# creating a Sequential machine learning model
model = Sequential()
model.add(Dense(128, input_shape=(len(train_x[0]), ),
                activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(len(train_y[0]), 
                activation='softmax'))

# compiling the model
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
              optimizer=sgd, metrics=['accuracy'])
hist = model.fit(np.array(train_x), np.array(train_y),
                 epochs=200, batch_size=5, verbose=1)

# saving the model
model.save("chatbotmodel.h5", hist)

# print statement to show the 
# successful training of the Chatbot model
print("Yay!")

Output:

Creating a main.py to run the Chatbot

We're done training the model, now we need to create the main file that will make the Chatbot model work and respond to our inputs.

Step 1: To get started, import the following modules.

Python3

# required modules
import random
import json
import pickle
import numpy as np
import nltk
from keras.models import load_model
from nltk.stem import WordNetLemmatizer

Step 2: Initialize the following classes and file contents.

Python3

lemmatizer = WordNetLemmatizer()

# loading the files we made previously
intents = json.loads(open("intense.json").read())
words = pickle.load(open('words.pkl', 'rb'))
classes = pickle.load(open('classes.pkl', 'rb'))
model = load_model('chatbotmodel.h5')

Step 3: We'll define 3 functions here.

clean_up_sentences(sentence) - This function will separate words from the sentences we'll give as input.

Python3

def clean_up_sentences(sentence):
    sentence_words = nltk.word_tokenize(sentence)
    sentence_words = [lemmatizer.lemmatize(word) 
                      for word in sentence_words]
    return sentence_words

bagw(sentence): This function will append 1 to a list variable 'bag' if the word is contained inside our input and is also present in the list of words created earlier.

First we'll use the function defined above to separate out 'root' words from the input, then in the next line, initialise a list variable called bag that will contain as many 0's as the length of the words list.
Using nested for loops, we'll check whether the word in the input is also in the words list. If it is, we'll append 1 to the bag, otherwise it'll remain 0.
Return a numpy array of the list variable bag that now contains 1's and 0's.

Python3

def bagw(sentence):
  
    # separate out words from the input sentence
    sentence_words = clean_up_sentences(sentence)
    bag = [0]*len(words)
    for w in sentence_words:
        for i, word in enumerate(words):

            # check whether the word
            # is present in the input as well
            if word == w:

                # as the list of words
                # created earlier.
                bag[i] = 1

    # return a numpy array
    return np.array(bag)

predict_class(sentence): This function will predict the class of the sentence input by the user.

Initialize a variable bow that will contain a NumPy array of 0's and 1's, using the function defined above. Using the predict() function, we'll predict the result based on the user's input.
Initialize a variable ERROR_THRESHOLD and append from 'res' if the value is greater than the ERROR_THRESHOLD, then sort it using the sort function.
Using a list variable return_list, store the tag or classes that was in the intense.json file.

Python3

def predict_class(sentence):
    bow = bagw(sentence)
    res = model.predict(np.array([bow]))[0]
    ERROR_THRESHOLD = 0.25
    results = [[i, r] for i, r in enumerate(res) 
               if r > ERROR_THRESHOLD]
    results.sort(key=lambda x: x[1], reverse=True)
    return_list = []
    for r in results:
        return_list.append({'intent': classes[r[0]],
                            'probability': str(r[1])})
        return return_list

get_response(intents_list, intents_json): This function will print a random response from whichever class the sentence/words input by the user belongs to.

First we initialize some required variables, such as tags, list_of_intents and results.
If the tag matches the tags in the list_of_intents, store a random response in a variable called result, using the choice() method of the random module.
Return result.

Python3

def get_response(intents_list, intents_json):
    tag = intents_list[0]['intent']
    list_of_intents = intents_json['intents']
    result = ""
    for i in list_of_intents:
        if i['tag'] == tag:
          
              # prints a random response
            result = random.choice(i['responses'])  
            break
    return result

print("Chatbot is up!")

Step 4: Finally, we'll initialize an infinite while loop that will prompt the user for an input and print the Chatbot's response.

Python3

while True:
    message = input("")
    ints = predict_class(message)
    res = get_response(ints, intents)
    print(res)

main.py is as follows:

Python3

import random
import json
import pickle
import numpy as np
import nltk
from keras.models import load_model
from nltk.stem import WordNetLemmatizer

lemmatizer = WordNetLemmatizer()
intents = json.loads(open("intense.json").read())
words = pickle.load(open('words.pkl', 'rb'))
classes = pickle.load(open('classes.pkl', 'rb'))
model = load_model('chatbotmodel.h5')

def clean_up_sentences(sentence):
    sentence_words = nltk.word_tokenize(sentence)
    sentence_words = [lemmatizer.lemmatize(word)
                      for word in sentence_words]
    return sentence_words

def bagw(sentence):
    sentence_words = clean_up_sentences(sentence)
    bag = [0]*len(words)
    for w in sentence_words:
        for i, word in enumerate(words):
            if word == w:
                bag[i] = 1
    return np.array(bag)

def predict_class(sentence):
    bow = bagw(sentence)
    res = model.predict(np.array([bow]))[0]
    ERROR_THRESHOLD = 0.25
    results = [[i, r] for i, r in enumerate(res)
               if r > ERROR_THRESHOLD]
    results.sort(key=lambda x: x[1], reverse=True)
    return_list = []
    for r in results:
        return_list.append({'intent': classes[r[0]],
                            'probability': str(r[1])})
        return return_list

def get_response(intents_list, intents_json):
    tag = intents_list[0]['intent']
    list_of_intents = intents_json['intents']
    result = ""
    for i in list_of_intents:
        if i['tag'] == tag:
            result = random.choice(i['responses'])
            break
    return result

print("Chatbot is up!")

while True:
    message = input("")
    ints = predict_class(message)
    res = get_response(ints, intents)
    print(res)

Output:

zehraina

Improve

Article Tags :

Deploy a Chatbot using TensorFlow in Python

Modules required:

Training the Chatbot model

Creating a main.py to run the Chatbot

Similar Reads

Deep Learning Basics

Neural Networks Basics

Deep Learning Models

Deep Learning Frameworks

Model Evaluation

Deep Learning Projects

Thank You!

What kind of Experience do you want to share?