ai-chatbot.ipynb

{"cells":[{"cell_type":"markdown","metadata":{},"source":["# <div style=\"background-color: #153B50; font-family: Trebuchet MS; color: #FFFFFF; padding: 5px; font-size: 150%; text-align: center; line-height: 1; border-radius: 10px 10px;\">AI ChatBot</div>\n","\n","Intelligent ChatBot built with Microsoft's DialoGPT transformer to interact with human users!"]},{"cell_type":"markdown","metadata":{},"source":["# <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">What's a ChatBot?</div>\n","\n","ChatBot is a kind of virtual assistant that can build conversations with human users! Robot Chatting. Building a chatbot is one of the popular tasks in Natural Language Processing (NLP).\n","\n","## <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">Are all chatbots the same?</div>\n","\n","Chatbots are classified into three categories:\n","1. Rule-based chatbots\n","2. Retrieval-based chatbots\n","3. Intelligent chatbots\n","\n","### <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">Rule-based chatbots</div>\n","\n","Rule-based chatbots are a type of chatbot that operates based on a predetermined set of rules or logic. They use these rules to process user input and provide responses according to predetermined patterns. Rule-based chatbots do not have the ability to deeply understand context or complex language, as they only respond based on pre-programmed rules.\n","\n","### <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">Retrieval-based chatbots</div>\n","\n","A retrieval-based chatbot selects responses from a pre-programmed list of answers based on similarities between user requests and available data. They employ pattern matching algorithms or choose the best answers based on keywords or other characteristics in user input. These chatbots do not generate new replies; rather, they select the best response from a predefined list of answers.\n","\n","### <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">Intelligent chatbots</div>\n","\n","Intelligent chatbots are a type of chatbot that uses sophisticated artificial intelligence technologies and algorithms, such as natural language processing (NLP), machine learning, and contextual understanding to understand and respond to user input in a more complex and adaptive way. Smart Chatbots can learn from user interaction and improve performance over time. They can understand more natural languages, deal with complex questions, and provide more contextual and relevant responses."]},{"cell_type":"markdown","metadata":{},"source":["## <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">In this notebook?</div>\n","\n","We will describe how we created an intelligent AI chatbot based on Microsoft's well-known transformer architecture, DialoGPT. DialoGPT, built on the Hugging Face card concept, is a cutting-edge generation response model for multi-turn talks. Human evaluation results revealed that the response created by DialogPT is comparable to the quality of human responses in the Turing single-turn conversation test. The model was trained using 147 million multi-turn interactions from Reddit's discussion thread."]},{"cell_type":"markdown","metadata":{},"source":["# <div style=\"font-family: Trebuchet MS; background-color: #153B50; color: #FFFFFF; padding: 12px; line-height: 1.5;\">Let's Start</div>\n","\n","Import the required libraries and frameworks!"]},{"cell_type":"code","execution_count":null,"metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.execute_input":"2024-04-12T08:53:17.372609Z","iopub.status.busy":"2024-04-12T08:53:17.372196Z","iopub.status.idle":"2024-04-12T08:53:24.426252Z","shell.execute_reply":"2024-04-12T08:53:24.424885Z","shell.execute_reply.started":"2024-04-12T08:53:17.372570Z"},"trusted":true},"outputs":[],"source":["import numpy as np\n","import time\n","from transformers import AutoModelForCausalLM, AutoTokenizer\n","import torch\n","from transformers.utils import logging\n","logging.get_logger(\"transformers\").setLevel(logging.ERROR)"]},{"cell_type":"markdown","metadata":{},"source":["### Download the DialoGPT model and tokenizer from Microsoft\n","The Hugging Face checkpoint for the model and tokenizer is \"microsoft/DialoGPT-medium\""]},{"cell_type":"code","execution_count":null,"metadata":{"execution":{"iopub.execute_input":"2024-04-12T08:54:25.103904Z","iopub.status.busy":"2024-04-12T08:54:25.102543Z","iopub.status.idle":"2024-04-12T08:54:35.085647Z","shell.execute_reply":"2024-04-12T08:54:35.084337Z","shell.execute_reply.started":"2024-04-12T08:54:25.103819Z"},"trusted":true},"outputs":[],"source":["# checkpoint \n","checkpoint = \"microsoft/DialoGPT-medium\"\n","# download and cache tokenizer\n","tokenizer = AutoTokenizer.from_pretrained(checkpoint, padding_size=\"left\")\n","# download and cache pre-trained model\n","model = AutoModelForCausalLM.from_pretrained(checkpoint)"]},{"cell_type":"markdown","metadata":{},"source":["### Class ChatBot"]},{"cell_type":"code","execution_count":null,"metadata":{"execution":{"iopub.execute_input":"2024-04-12T08:58:15.026161Z","iopub.status.busy":"2024-04-12T08:58:15.025509Z","iopub.status.idle":"2024-04-12T08:58:15.045417Z","shell.execute_reply":"2024-04-12T08:58:15.044083Z","shell.execute_reply.started":"2024-04-12T08:58:15.026113Z"},"trusted":true},"outputs":[],"source":["# build a ChatBot class with all the modules needed to conduct a complete conversation\n","class ChatBot():\n","    # initialization\n","    def __init__(self):\n","        # once a conversation is started, history will be saved for smooth conversation\n","        self.chat_history_ids = None\n","        # makes input ids global for use anywhere in the object\n","        self.bot_input_ids = None\n","        # a sign to check whether to end the conversation\n","        self.end_chat = False\n","        # greeting text when starting a conversation\n","        self.welcome()\n","        \n","    def welcome(self):\n","        print(\"Initializing ChatBot ...\")\n","        # time to set up the user\n","        time.sleep(2)\n","        print('Type \"bye\" or \"quit\" or \"exit\" to end chat \\n')\n","        # allow time to read what has been printed\n","        time.sleep(3)\n","        # greetings and introductions\n","        greeting = np.random.choice([\n","            \"Welcome, I am ChatBot, here for your kind service\",\n","            \"Hey, Great day! I am your virtual assistant\",\n","            \"Hello, it's my pleasure meeting you\",\n","            \"Hi, I am a ChatBot. Let's chat!\"\n","        ])\n","        print(\"ChatBot >>  \" + greeting)\n","        \n","    def user_input(self):\n","        # receive input from the user\n","        text = input(\"User    >> \")\n","        # end the conversation if the user wishes\n","        if text.lower().strip() in ['bye', 'quit', 'exit']:\n","            # turn on the sign\n","            self.end_chat=True\n","            # closing remarks\n","            print('ChatBot >>  See you soon! Bye!')\n","            time.sleep(1)\n","            print('\\nQuitting ChatBot ...')\n","        else:\n","            # continuing the conversation, processing the input text\n","            # encode new user input, add eos_token, and return tensor in Pytorch\n","            self.new_user_input_ids = tokenizer.encode(text + tokenizer.eos_token, \\\n","                                                       return_tensors='pt')\n","\n","    def bot_response(self):\n","        # add new user input token to conversation history\n","        # jika percakapan sudah dimulai sebelumnya\n","        if self.chat_history_ids is not None:\n","            self.bot_input_ids = torch.cat([self.chat_history_ids, self.new_user_input_ids], dim=-1) \n","        else:\n","            # if first entry, initialize bot_input_ids\n","            self.bot_input_ids = self.new_user_input_ids\n","        \n","        # define new chat_history_ids based on previous conversations\n","        # generate responses while limiting total conversation history to 1000 tokens\n","        self.chat_history_ids = model.generate(self.bot_input_ids, max_length=1000, \\\n","                                               pad_token_id=tokenizer.eos_token_id)\n","            \n","        # the last output token of the bot\n","        response = tokenizer.decode(self.chat_history_ids[:, self.bot_input_ids.shape[-1]:][0], \\\n","                               skip_special_tokens=True)\n","        # if the bot fails to answer\n","        if response == \"\":\n","            response = self.random_response()\n","        # show bot response\n","        print('ChatBot >>  '+ response)\n","        \n","    # if there is no response from the model\n","    def random_response(self):\n","        i = -1\n","        response = tokenizer.decode(self.chat_history_ids[:, self.bot_input_ids.shape[i]:][0], \\\n","                               skip_special_tokens=True)\n","        # iterate backwards over the history to find the last token\n","        while response == '':\n","            i = i-1\n","            response = tokenizer.decode(self.chat_history_ids[:, self.bot_input_ids.shape[i]:][0], \\\n","                               skip_special_tokens=True)\n","        # If it's a question, answer it appropriately\n","        if response.strip() == '?':\n","            reply = np.random.choice([\"I don't know\", \n","                                     \"I am not sure\"])\n","        # not a question? answer correctly\n","        else:\n","            reply = np.random.choice([\"Great\", \n","                                      \"Fine. What's up?\", \n","                                      \"Okay\"\n","                                     ])\n","        return reply"]},{"cell_type":"markdown","metadata":{},"source":["### Happy Chatting"]},{"cell_type":"code","execution_count":null,"metadata":{"execution":{"iopub.execute_input":"2024-04-12T08:58:53.775505Z","iopub.status.busy":"2024-04-12T08:58:53.774982Z","iopub.status.idle":"2024-04-12T09:00:57.603284Z","shell.execute_reply":"2024-04-12T09:00:57.601817Z","shell.execute_reply.started":"2024-04-12T08:58:53.775467Z"},"trusted":true},"outputs":[],"source":["# build obyek ChatBot\n","bot = ChatBot()\n","# start chatting\n","while True:\n","    # get user input\n","    bot.user_input()\n","    # if you want to end the conversation\n","    if bot.end_chat:\n","        break\n","    # output bot respons\n","    bot.bot_response()"]},{"attachments":{"6134927b-8ce6-4880-9481-666c1761b879.png":{"image/png":"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"}},"cell_type":"markdown","metadata":{},"source":["### Sampel chat\n","![image.png](attachment:6134927b-8ce6-4880-9481-666c1761b879.png)"]},{"cell_type":"markdown","metadata":{},"source":["*Thank you for visit*"]}],"metadata":{"kaggle":{"accelerator":"none","dataSources":[],"dockerImageVersionId":30646,"isGpuEnabled":false,"isInternetEnabled":true,"language":"python","sourceType":"notebook"},"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.12.3"}},"nbformat":4,"nbformat_minor":4}