AI+ Tool and Solution Catalogue

- Our AI-Powered Speech Recognition & Analytics Platform is an enterprise-grade solution designed to automate the conversion of speech to accurate text, tailored for the linguistic complexities of Hong Kong. - It addresses the critical business need of reducing the time and cost associated with manual transcription, thereby improving operational efficiency and public service delivery. - Leveraging state-of-the-art AI models, our platform achieves market-leading accuracy (Word Error Rate below 5%) for Cantonese, English, Mandarin, and mixed-language audio. - The solution offers both real-time "live" transcription for immediate applications and high-throughput "batch" processing for large volumes of recordings. - It is designed for secure on-premise deployment to meet stringent government data security requirements and can be enhanced with Generative AI capabilities for automated summarization and insight extraction.

The solution empowers government departments to transform audio data into structured, actionable text, significantly improving public services and internal workflows.

How it improves public services: - Live Transcription: Provides real-time captions for public-facing counters, interviews, and press conferences, enhancing accessibility for the hearing-impaired and creating immediate records. For example, a department can offer instant voice-to-text services for citizens at a service counter. - Batch Transcription: Efficiently processes large backlogs of audio recordings from sources like public hearings, inquiries, or evidence files. This accelerates investigation processes, policy review, and information disclosure. - Custom Dictionary: Departments can build and manage their own dictionaries with domain-specific terminology (e.g., legal terms, policy names, technical jargon). This ensures unparalleled accuracy for specialized content, leading to higher quality public-facing documents and records.

Costs and Benefits: - Drastic Efficiency Gain: Reduces transcription time from hours to minutes, freeing up staff from laborious manual work to focus on high-value tasks. Our benchmark shows processing speeds of up to half the audio length. - Enhanced Accuracy & Consistency: AI models provide consistent, high-accuracy transcripts, minimizing human error, especially for complex or mixed-language content. - Improved Data Accessibility & Searchability: Converts unstructured audio into searchable text, allowing departments to quickly locate key information within vast audio archives. - Actionable Insights: By integrating with Generative AI, the platform can automatically summarize long meetings, identify key topics, and extract action items, turning raw data into strategic insights. - Full Data Sovereignty: The on-premise deployment model ensures all sensitive audio and text data remains within the government's secure infrastructure. - Costs: The primary costs are the initial software license and system integration, followed by an annual maintenance and support fee. This investment is quickly offset by significant savings in man-hours and outsourcing costs previously spent on manual transcription.