Challenge
At the heart of Taiwan’s bustling cities, the Mass Rapid Transit (MRT) stations serve as vital hubs, sustaining the nation's dynamic economic landscape and fostering productivity by facilitating the seamless movement of people worldwide. Catering to both residents and international visitors, the MRT offers convenient services for efficient and smooth transportation. Despite these strengths, current multilingual language support of on-site customer service is limited, resulting in communication barriers and inefficiencies. This limitation hampers Taiwan’s potential for global exposure through socio-economic development and tourism, hindering progress in regional integration, investment attraction, overseas tourism, and more.
Taiwan grapples with a dual challenge involving a shortage of multilingual talent and rising labor costs. The complexity is compounded by the need for proficiency in multiple languages concurrently, leading employers to enlist several language specialists simultaneously, causing a substantial increase in human resource expenses. Providing MRT employees with multilingual proficiency for bilingual customer services becomes a protracted and potentially costly endeavor. Furthermore, the challenge of securing ad-hoc sign-language support remains prevalent in most MRT stations, leaving inclusivity challenges unresolved for minority groups.
In response, AI ChatBots have begun integrating into customer service and support. However, the current landscape of AI voice customer service systems utilizing Large Language Models (LLMs) poses substantial challenges in terms of computing power when operating on edge systems for real-time multilingual translation and communication. The majority of AI LLMs are exceptionally large, depending on the robust computing capabilities of cloud servers rather than edge computing. This reliance on cloud computing renders it impractical to execute offline, high-speed, real-time processing on local machines while ensuring the utmost security and privacy.
Solution
The progression of intelligent MRT customer service demands a shift towards unmanned technology. Introducing the AI voice and visual customer service system, this innovative solution integrates dynamic displays featuring virtual characters/ avatars that closely emulate real-life interactions, providing an interactive approach to answering passenger inquiries regarding basic station information, ticketing, surrounding transportation, and attractions. ASRock Industrial's iEPF-10000S Series Robust Edge AIoT Platform functions as an AI inference edge system equipped with Intel's 14th/13th Gen CPU and NVIDIA GeForce RTX® 4090 graphics card, featuring high-performance CPU, GPU and AI computing power at the edge.
iEPF-10000S Series Robust Edge AIoT Platform features Intel's 13th Gen CPU as an advanced AI inference edge system utilizing LLM and ASR speech-to-text technology for enhanced AI ChatBot’s multilingual support
This comprehensive system offers simultaneous multilingual support, functioning offline with integrated pre-trained Large Language Model (LLM) models. It excels in providing simultaneous translation in 14 languages (English, Chinese, Japanese, Korean, Thai, Indonesian, Malay, Spanish, French, Russian, Italian, German, Dutch, and Portuguese), along with country-specific sign language support, while also allowing for potential expansion to meet additional language demands. The system ensures swift, zero-lag, and accurate customer service by leveraging a combination of two multimodal translation AIs, guaranteeing the highest quality translations. This multilingual capability provides a seamless and inclusive communication experience for passengers, enhancing the efficiency, quality, and accessibility of MRT customer service.
Supporting Automatic Speech Recognition (ASR) and speech-to-text technology, the system efficiently eliminates language barriers, promoting smooth global communication with enhanced efficiency. Continuously evolving, it remains committed to learning and optimizing AI models for updated knowledge and language support through the AI ChatBot. The execution encompasses vital processes: language input, language text conversion, AI semantic understanding, searching for answers to questions, and integrating virtual avatars with synthesized voice responses. Furthermore, operated locally, the system ensures internet-free functionality with secure privacy features, providing dependable translation services both online and offline without compromising data security.