Yermek D.2025-06-132025-06-132023Yermek D / Creation of effective chatbots based on neural networks / 7M05401 - Department of Engineering / 2023https://repository.sdu.edu.kz/handle/123456789/1762This paper presents a comprehensive exploration of the development process for a neural network-based chatbot. Given the increasing popularity of neural networks and chatbots in recent years, investigating the creation process of such a chatbot is both pertinent and valuable. The primary objective of this study is twofold: first, to construct an effective chatbot, and second, to identify the most precise model for text classification. Furthermore, the paper delves into the mathematical aspects of creating a chatbot and provides a detailed explanation of neural networks. The initial step in our project involved the creation of a database containing a sufficient number of student questions, which were then categorized into ten distinct categories. ‘To achieve this, we carefully selected the ten most commonly asked questions and generated multiple paraphrased versions of each question. These variations were employed for training and evaluating our models. During the course of our research, we identified three text classification models that proved to be the most suitable for our purposes: Multinomial Logistic Reeression, Naive Bayes, and Neural Network. We conducted extensive tests using these models and documented the results in a table, providing a comprehensive analysis of their performance. Following the successful completion of the database collection and the identification of the optimal text classification model, we proceeded to create the chatbot using DialogF low. Additionally, we integrated our chatbot into the Telegram messenger for wider accessibility and user convenience.enThesis