View Proposal

---

Proposer

Kah Kit Ng

Title

C.A.M.P.U.S. (China-Adapting Malaysian Platform for University Students): A Centralized AI Dashboard for Cultural and Digital Orientation

Goal

1. Develop a Multilingual Retrieval-Augmented Generation (RAG) Pipeline: Engineer a system capable of querying a local vector database of English and Malay documents using conversational Mandarin prompts, specifically leveraging DeepSeek's cross-lingual reasoning capabilities. 2. Design a Dual-Facing Web Platform: Construct a personalized student dashboard ("My Local Life") for managing adaptation progress and an administrative analytics dashboard that visualizes trending queries and system engagement. 3. Evaluate System Performance and User Acceptance: Benchmark the technical efficiency of the local LLM deployment (latency and retrieval accuracy) and evaluate user adoption using the Technology Acceptance Model (TAM) through statistical analysis.

Description

Existing translation tools perform poorly on highly localized, culturally specific Malaysian contexts, such as navigating a Mamak stall menu or understanding campus-specific administrative terminology. While Large Language Models (LLMs) offer conversational assistance, foundation models lack hyper-local, up-to-date knowledge and are prone to hallucinating administrative procedures. Consequently, Chinese international students experience prolonged acculturative stress, while university International Offices lack real-time visibility into the specific challenges new cohorts face. There is a need for a centralized, context-aware platform that retrieves accurate local information, maps it culturally into Mandarin, and provides actionable analytics to university administrators.

Resources

Background

International students from China transitioning to Malaysian universities face a unique dual-challenge: a traditional linguistic barrier combined with a severe "digital ecosystem shock." The immediate shift from familiar, ubiquitous platforms (WeChat, Alipay, Meituan) to highly localized applications (WhatsApp, Grab, Touch 'n Go) creates significant friction. Furthermore, navigating local university procedures, colloquial slang (Manglish), and diverse dietary options requires cultural context that standard translation applications cannot provide. Information critical to this transition remains highly fragmented across English and Malay university portals, government websites, and student forums.

Url

Difficulty Level

Moderate

Ethical Approval

InterfaceOnly

Number Of Students

1

Supervisor

Kah Kit Ng

Keywords

artificial intelligence, human-computer interaction (hci), large language models (llms)

Degrees

Bachelor of Science in Computing Science