IMPROVING THE KALAKA WEBSITE WITH SVELTE FRAMEWORK AND SPEAKER DIARIZATION
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak
Abstract
Description
The Kalaka system, developed by UNIMAS in 2017, is an innovative platform aimed at preserving Malaysia’s rich linguistic and cultural heritage (Juan & Busu, 2017). It serves as a valuable tool for studying and documenting the voices of Malaysia’s ethnic communities. To expand its capabilities, Kalaka needs AI-driven tools like speaker diarization, which can unlock new opportunities for linguistic research. This project builds upon the work of Mohd Zulhafiz Rahim, Dr. Sarah Flora Samson Juan, and Professor Dr. Fitri Suraya Mohamad (2023), who focused on improving speaker diarization for low-resourced Sarawak Malay speech using advanced clustering techniques and x vectors. Speaker diarization is a powerful technology that answers the question “Who spoke when?” by detecting and segmenting different voices within an audio recording (Sayyada et al.). It improves transcription accuracy and allows for in-depth speaker analysis, making it a valuable tool for researchers studying multilingual conversations. However, the current Kalaka system has several limitations. Its frontend is outdated and lacks interactivity, resulting in low user engagement. The system only offers basic features and does not support real-time updates or data visualizations. There is no dedicated language page to promote cultural learning, and advanced tools like speaker diarization are not yet integrated, limiting its effectiveness for audio analysis and language documentation. Furthermore, there is a need to enrich the platform with educational and engaging features such as a Word of the Day (WOTD), an interactive Glossary for exploring vocabulary, and Quiz modules to test user knowledge and promote active learning. To address these issues, this project has three main objectives: to review existing systems for feature ideas, to design and develop an improved version of Kalaka using the Svelte framework with speaker diarization integration, and to evaluate the new system through developer and user acceptance testing. The Prototyping Model will be used as the development methodology, emphasizing iterative development and continuous feedback. This approach ensures the project remains user-focused, allowing early prototypes to inform refinements and lead to a more effective final product. In conclusion, with its new revamp, the Kalaka system is expected to enhance user engagement through improved features and functionality. This chapter has demonstrated that the objectives of the proposed system have been successfully achieved, while its limitations have also been identified. The proposed future enhancements aim to address these limitations and further refine the system. Implementing these improvements will be essential to ensure a more effective and seamless user experience in real-world applications.