Hasbul Ihza Firnanda Az, . (2026) PENERAPAN LLM, FEM, DAN OPENAI WHISPER UNTUK UMPAN BALIK MULTIMODAL DALAM PELATIHAN WAWANCARA BERBASIS WEBSITE. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.
![[img]](http://repository.upnvj.ac.id/style/images/fileicons/text.png) |
Text
ABSTRAK.pdf Download (203kB) |
|
Text
AWAL.pdf Download (857kB) |
|
Text
BAB 1.pdf Restricted to Repository UPNVJ Only Download (183kB) |
|
Text
BAB 2.pdf Restricted to Repository UPNVJ Only Download (443kB) |
|
Text
BAB 3.pdf Restricted to Repository UPNVJ Only Download (1MB) |
|
Text
BAB 4.pdf Restricted to Repository UPNVJ Only Download (12MB) |
|
Text
BAB 5.pdf Download (169kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (171kB) |
|
Text
RIWAYAT HIDUP.pdf Restricted to Repository UPNVJ Only Download (482kB) |
|
Text
LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (849kB) |
|
Text
HASIL PLAGIARISME.pdf Restricted to Repository staff only Download (29MB) |
|
Text
ARTIKEL KI.pdf Restricted to Repository staff only Download (199kB) |
Abstract
Interview preparation remains a common challenge for job seekers, particularly in delivering answers clearly, confidently, and in accordance with the given questions. This study develops a web-based interview training platform that allows users to practice independently and receive direct feedback after each session. The system applies a Large Language Model (LLM), Facial Expression Model (FEM), and OpenAI Whisper to provide multimodal feedback covering both verbal and nonverbal aspects. OpenAI Whisper is used to transcribe users’ spoken answers into text, FEM is used to analyze facial expressions during the interview session, while the LLM is used to generate interview questions and evaluate the quality of users’ answers. The developed platform includes several main features, such as user authentication, profile management, interview session creation, answer recording, evaluation result display, and session history. The testing results show that all 15 Black-Box testing scenarios were successfully executed. The ASR model processed 687 audio data with a cleaned WER of 0.0865, a cleaned CER of 0.0306, and an average latency of 0.4262 seconds per file. The FEM model achieved an accuracy of 93.2%, while the LLM evaluation showed that DeepSeek V4 Flash was more suitable for question generation and Gemini 2.5 Flash was more suitable for answer evaluation. The User Acceptance Testing results also indicate that the system falls into the very good category, making it suitable to be used as a web-based interview training medium.
| Item Type: | Thesis (Skripsi) |
|---|---|
| Additional Information: | [No.Panggil: 2210511124] [Pembimbing: Noor Falih] [Penguji 1: Indra Permana Solihin] [Penguji 2: Kharisma Wiati Gusti] |
| Uncontrolled Keywords: | Interview Training, Multimodal Feedback, Large Language Model, OpenAI Whisper, Facial Expression Model |
| Subjects: | Q Science > QA Mathematics > QA76 Computer software |
| Divisions: | Fakultas Ilmu Komputer > Program Studi Informatika (S1) |
| Depositing User: | HASBUL IHZA FIRNANDA AZ |
| Date Deposited: | 28 Jul 2026 03:16 |
| Last Modified: | 28 Jul 2026 03:17 |
| URI: | http://repository.upnvj.ac.id/id/eprint/51838 |
Actions (login required)
 |
View Item |