AI-Chatbot-with-Tensorflow

This is a Chatbot which i created for my final year engineering project. It is fully written in python and for training and modelling i have used Tensorflow. finally GUI is created with Tkinter in python my goal was to make this project more real world. Chatbot can be integrated with web too, Using flask nd tensorflow serving, but i haven't done that yet.

Working

So, the flow of the chatbot is as follows:

1. First we have a file called 'intents.json', in this file we have a set of predefined questions and answers. All questions and answers can be seperated by 'tags'. 'tags' here means a specific topic or category of a question. For ex, tag=Canteen means the question asked by the user falls into canteen category.
2. You can easily add or delete questions/answers into/from the intents.json file.
3. We have a neural network which will train on the questions, obviously we need to convert text into bag-of-words representation in order to do that. Here, we will learn the association between the question and the tag. Which means our feature variable is Bag of words representation of the questions and our target varaible is the tag associated with those questions. Yes, it's a classification task.
4. Once we find the tag, we just need to return an answer for that tag entry through the json file.

Example:

   {
                        "tag": "goodbye",
                        "patterns": [
                                "cya",
                                "see you",
                                "bye bye",
                                "See you later",
                                "Goodbye",
                                "I am Leaving",
                                "Bye",
                                "Have a Good day",
                                "talk to you later",
                                "tyyl",
                                "i got to go",
                                "gtg"
                        ],
                        "responses": [
                                "Sad to see you go :(",
                                "Talk to you later",
                                "Goodbye!",
                                "Come back soon"
                        ],
                        "context_set": ""
                },

So as you can see here, for tag='goodbye' people can say it in multiple ways, so for a single tag goodbye their are multiple questions, and also if our chatbot only answers one single answer then it would be so boring, that's why we have multiple answers to the goodbye tag, and each time it will choose randomly from that set.
User Enters --> See you / see you later/ bye --> bot predicts tag='goodbye' --> inside intents.json looks for tag='goodbye' return random response from responses list. So that's the flow.

Usage:

Clone the repo:

git clone https://github.com/Sanket758/AI-Chatbot-with-Tensorflow.git
cd AI-Chatbot-with-Tensorflow

Install requirements first

pip install -r requirements.txt

*You may need to download some nltk libraries manually.

    import nltk  
    nltk.download('punkt')  
    nltk.download('stopwords')  
    nltk.download('wordnet')

otherwise run -
python setup.py

Make sure that you have the model and intents file available in your directory.

For the Command-Line Version:

python Chatbot_v2_CLI.py

For GUI version use (uses Tkinter-Python):

python Chatbot-GUI.py

For the Web version (uses Flask-python):

python flask_app.py

Watch the demo here: