Development of student intent-based educational chatbot system with adaptive and attentive DTCN on symmetric convolution approach

• Published in: MethodsX Vol. 15; p. 103542
• Main Authors: Kathole, Atul, Patil, Suvarna, Jadhav, Dr. Devyani, Pathak, Hirkani, Mirge, Amita Sanjiv
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.12.2025; Elsevier
• Subjects: Adaptive and attentive deep temporal convolutional network with symmetric convolution; Averaging-based driving training-barnacles mating optimizer; Bidirectional encoder representations from transformers; Chatbots; Computer Science; Intent classification; Text convolutional neural network; Transformer network; Intent classification; Bidirectional encoder representations from transformers; Text convolutional neural network; Chatbots; Averaging-based driving training-barnacles mating optimizer; Transformer network; Adaptive and attentive deep temporal convolutional network with symmetric convolution
• ISSN: 2215-0161; 2215-0161
• DOI: 10.1016/j.mex.2025.103542

The Accessing academic information like exam time tables, exam scores, syllabus and tutor information on institutional websites can be time consuming and tedious on behalf of the student. In this case, we will solve this dilemma by creating an intelligent and an automated Student Intent-based Educational Chatbot that integrates deep learning along with state-of-art optimization algorithms. Following the pre-processing of student inquiries with generic chatbot data sets, the system starts operating. The text data that were cleaned are represented as vectors through three high-performing language representations, i.e., BERT, TransformerNet, and Text CNN. Such vectors are subject to weighted selection of features and fusion where best features are selected and fused together by Averaging-based Driving Training -Barnacles Mating Optimizer (ADT -BMO) is a new hybrid metaheuristic optimization algorithm. ADT-BMO is smart when it comes to weighting the feature and optimization parameter to maximize the relevance of fused features. Thicken with Symmetric Convolution (AA-DTCN-SC) and the down-stream adaptive feature set refined above are fed to this network to achieve accurate intent recognition. ADT-BMO also improves AA-DTCN-SC model by optimizing the parameters hidden neurons, activation functions and epochs under which the classification accuracy is high. In testing the noted purpose, there is an automatic building of system responses to queries formulated using contextually right answers. The experimental simulations also illustrate the effectiveness of the superiority of this approach as the chatbot performs better than its baseline models of DTCN, RNN and Bi-LSTM by 4.44%, 3.3%, 10.59%, 11.9, respectively. The given research therefore presents a well-functioning, scalable and time-saving educational chatbot, which improves student engagement through provision of quick, precise, and pertinent scholarly assistance.•To help save the time and work required of students and administrators by replying to common academic questions through the creation of an automated intelligent chatbot that is able to find correct and instant answers based on deep learning and NLP approaches.•In order to eliminate the shortcomings of current chatbot platforms, i.e. irrelevant outputs and fabricated predictivity of user intent, through implementation of a different novel model (AA-DTCN-SC) on deep learning, which is optimized using a hybrid ADT-BMO algorithm to aid intent inference and user interaction in educational settings. Illustration of the generated Student Intent-based Educational Chatbot [Display omitted]