Pembuatan Sistem Visual Question Answering Berbasis Web Untuk Mendukung Pembelajaran Visual Anak TK Berbahasa Indonesia Menggunakan Deep Learning

Hanifah, Asiyah (2023) Pembuatan Sistem Visual Question Answering Berbasis Web Untuk Mendukung Pembelajaran Visual Anak TK Berbahasa Indonesia Menggunakan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05311940000002-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (2MB) | Request a copy

Abstract

Seiring pesatnya perkembangan teknologi, Indonesia semakin gencar melakukan persiapan transformasi digital untuk menghadapi perubahan teknologi. Salah satunya adalah implementasi e-learning di berbagai sektor, termasuk pendidikan. E-learning telah diterapkan dalam pembelajaran taman kanak-kanak, termasuk pembelajaran visual. Bentuk pembelajaran visual pada e-learning dapat dibuat dengan pembangunan sistem visual question answering. Beberapa penelitian telah dibuat untuk pembangunan sistem visual question answering dan berhasil membuat sistem visual question answering dengan ilmu patologi dalam bahasa inggris, dan dataset objek di sekitar monas dalam bahasa indonesia. Oleh karena itu, dilakukan pengajuan pembuatan sistem visual question answering dengan dataset yang lebih umum dan dapat dikenali oleh anak TK berbahasa indonesia. Dadanya penelitian ini akan dapat membantu tenaga pendidik dalam kegiatan belajar mengajar yang lebih interaktif dalam e-learning. Penelitian ini menggunakan model Bootstrapping Language-Image Pre-training (BLIP) untuk proses pembuatan sistem visual question answering dan mengimplementasikan model No Language Left Behind (NLLB) pada input-output pertanyaan untuk menerjemahkan bahasa yang digunakan. Hasil implementasi kedua model BLIP dan NLLB berhasil membangun sistem visual question answering berbahasa indonesia. Berdasarkan hasil pengujiannya, dari beberapa pertanyaan yang mengandung 6 jenis jawaban: ya/tidak, kata benda, kata kerja, kata sifat, kata keterangan, dan numeral, sistem ini berhasil menjawab tepat untuk jenis jawaban ya/tidak, kata benda, kata kerja, dan kata keterangan dengan nilai ketepatan jawaban ya/tidak 100, kata benda 100, kata kerja 100, dan kata keterangan 87.5.
===============================================================================================================================
Along with the rapid development of technology, Indonesia is increasingly intensifying its digital transformation preparations to face the existing technological advancements. With the rapid development of technology, Indonesia is increasingly intensifying its preparations for digital transformation to face technological changes. One of these preparations involves the implementation of e-learning across various sectors, including education. E-learning has been applied in kindergarten education, including visual learning. Visual learning in e-learning can be achieved through the development of a visual question answering system. Several studies have been conducted on visual question answering systems and have successfully created such systems using pathology images in English and object datasets around Monas (National Monument) in Indonesian. Therefore, the author proposes the creation of a visual question answering system with a more general dataset that can be recognized by Indonesian-speaking kindergarten children. The purpose of this research is to assist educators in conducting more interactive e-learning activities. The research utilizes the Bootstrapping Language-Image Pre-training (BLIP) model for the development of the visual question answering system and implements the No Language Left Behind (NLLB) model for translating the language used in the input-output questions. The implementation results of both the BLIP and NLLB models successfully build a visual question answering system in the Indonesian language. Based on testing, the system can provide accurate answers for yes/no, nouns, verbs, and adverbial questions, with accuracy rates of 100% for yes/no, 100% for nouns, 100% for verbs, and 87.5% for adverbial questions, all of which contain six types of answers: yes/no, nouns, verbs, adjectives, adverbs, and numerals.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	E-learning, Sistem Visual Question Answering, Bootstrapping Language-Image Pre-training, No Language Left Behind
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Asiyah Hanifah
Date Deposited:	01 Sep 2023 07:37
Last Modified:	01 Sep 2023 07:37
URI:	http://repository.its.ac.id/id/eprint/102392

Actions (login required)

View Item