It is demonstrated that ANN-based intent classification can effectively reduce reliance on manual library staff for routine inquiries while providing continuous 24/7 automated support in a university digital library setting.
Abstract
Academic libraries increasingly struggle to provide consistent information services as digital collections grow and user inquiries become more varied. Staff-dependent communication channels such as email and live chat are constrained by working hours, leaving visitors without support during evenings, weekends, and holidays. This study proposes an intelligent text-based chatbot to automate visitor inquiry handling in a university digital library setting. The system employs an Artificial Neural Network (ANN) with a multi-layer Perceptron architecture trained on a corpus of 2,319 labeled samples across 56 intent categories. An Indonesian-language preprocessing pipeline was implemented, consisting of case folding, tokenization, stopword removal, and morphological stemming using the Sastrawi algorithm, followed by Bag-of-Words feature extraction producing 1,494-dimensional vectors. The system was deployed using a decoupled architecture, separating the Python Flask inference backend from a PHP CodeIgniter 4 frontend via an Ngrok HTTPS tunnel. Using an 80:20 stratified train-test split, the model achieved a classification accuracy of 91.00%, with a macro-average precision of 0.90 and recall of 0.92 on 464 test samples. Black-box functional testing confirmed stable real-time performance with response latency under three seconds. These results demonstrate that ANN-based intent classification can effectively reduce reliance on manual library staff for routine inquiries while providing continuous 24/7 automated support.
This paper presents the design and implementation of intelligent chatbots using Retrieval-Augmented Generation architecture to derive accurate, context-based responses from domain-specific knowledge sources such as PDFs and CSV files. The proposed system integrates information retrieval mechanisms with large language models to overcome crucial limitations of traditional chatbots: hallucination, knowledge cut-off, and inability to access private organizational data. The proposed architecture uses a six-stage pipeline consisting of document ingestion, text splitting, and embedding generation with sentence transformers, semantic retrieval, and response generation via Large Language Models (LLMs). By preprocessing the PDF and CSV documents using a RecursiveCharacterTextSplitter and converting them into 384-dimensional semantic embeddings, the system works via a similarity-based retrieval of document chunks. For implementation, the LlamaIndex framework was used for data indexing and query orchestration, and HuggingFace Transformers for language modeling. Experimental validation conducted on the smart campus chatbot system demonstrated significant improvements in the accuracy, relevance and transparency of responses across education, student services, and administrative support domains compared with a standalone language model. Because of its modular design, it supports a wide range of domain-specific topics without the combinatorial costs of re-training. This is another important application for systems that require regular updates and accurate information. Overall, the proposed framework provides a simple and practical approach for easily implementing domain-aware chatbots.
Varsha Spoorthi Shree Narravula, V. K, Yashaswini B. Suresh et al.· 2026 7th International Confe...· 0 citations
The study successfully validated that the intelligent chatbot efficiently bridges communication barriers, automates repetitive administrative inquiries, and improves service accessibility, confirming that the application is highly usable, practical, and effective as an inclusive, assistive communication tool.
Gil G. Dialogo, Hershey Alburo-Abugadie, Enrico C. Lucero· International journal of com...· 0 citations
Property information services at Griya Alam Mandiri currently rely on manual WhatsApp interactions during working hours, often leading to delayed responses for prospective buyers. This study aims to implement a Natural Language Processing (NLP) based chatbot using the IndoBERT model on the housing website to automate 24/7 customer service. The methodology involves extracting 3,082 training data samples augmented with typos and informal language, model fine-tuning, and benchmarking against comparative architectures. The evaluation results show that the IndoBERT Base model highly excels in intent classification with an accuracy of 95.85% and an F1-Score of 95.87%, outperforming Indonesian RoBERTa, IndoBERTweet, and XLM-RoBERTa. This system was successfully integrated into a Client-Server architecture connecting a React interface and a FastAPI backend. The final evaluation through User Acceptance Testing (UAT) involving 50 respondents achieved an overall feasibility percentage of 82.50% (Very Satisfied). In conclusion, the IndoBERT chatbot is proven effective in improving the speed of real-time information services and successfully reduces administrative workloads. Future research suggestions include integrating Large Language Models (LLM) with the Retrieval-Augmented Generation (RAG) method to generate more dynamic responses.
The system offers a replicable instantiation of RAG for transparent, citation-backed campus information services in Indonesian and offers a replicable instantiation of RAG for transparent, citation-backed campus information services in Indonesian.
Haddad Alwi Yafie, A. Zailani, Widang Muttaqin et al.· SinkrOn· 0 citations
The study concludes that AI-powered chatbot systems offer a substantive remedy to institutional communication inefficiencies and recommends full deployment alongside continuous model retraining using current institutional data.
Ulysses Chinenyeaka Ebizie· International Journal of Com...· 0 citations
It is concluded that an NLP- and RAG-based chatbot is feasible to be implemented as a digital campus information service and improves answer accuracy from 58.5% to 89.25%, with an average response time below three seconds.
Zuhri Yanda, Yeni Yanti, Maulinda et al.· IC-ITECHS· 0 citations