Chapter 3 Design • Train the model: The training data set is identified and used to prepare a model. • Test the model: This is a new dataset, different than the training set; you gather predictions from the trained model with the inputs from the test dataset and compare them with the withheld output values of the test set. • Improve and tune the model: You can adjust various parameters and tune the weights to improve the model built. Figure 3-27. AI automation model Chatbot Case Study Most common AI solutions are built using Tensorflow, Python, and Spark. AI strategy helps to solve defined business problems, with a defined data set to solve the problems. Chatbots are programs built with NLP, which is supposed to solve domain-specific problems and query request by simulating human conversation. There are three components: presentation layer; bot layer, which has the bot framework or engine; and transaction and data processing system, which interact with the DXP to integrate the existing system with the chatbot engine, as shown in Figure 3-28. • Presentation layer: The chat interface can be a custom UI (e.g., Angular, React) and native mobile application. You can interact with this layer using text, voice, and visual. Inputs are sent to the backend layer using web socket communication and REST APIs. • Bot engine: This is an open-source bot framework used to create the bot model for specific use cases and domains to understand the intent of the user. The bot framework has capabilities to process, understand, and generate language that is NLP, NLU, and NLG as follows. • NLP (natural language processing): This component understands the text or voice and understands the intent of the user. 107