Automated Sentiment Analysis System Decodes Dynamics of the Catering Industry

The Census and Statistics Department (C&SD) regularly conducts the Quarterly Survey of Restaurant Receipts and Purchases to measure the level and trends of consumer spending on catering services in Hong Kong. To stay abreast of market dynamics in a timely manner, statistical staff need to review, consolidate, and interpret a substantial volume of news reports. However, with the increasing diversification of media channels and the surge in information flow, traditional manual review processes have become increasingly time-consuming and are struggling to cope with the growing demands of data processing.

To address this challenge, C&SD is developing an automated sentiment analysis intelligent system powered by Large Language Models (LLMs) to enhance the efficiency of monitoring market dynamics. The system can automatically aggregate news from multiple media sources, filter content related to catering business operations, classify articles by restaurant type, and assess their sentiment orientation. Through these data-driven insights, C&SD can, along with capturing industry trends more promptly, support real-time assessment and now-casting of economic prospects, thereby improving the timeliness of relevant indicators and quality of official statistics.

Looking ahead, C&SD will continue to refine the system's performance and enhance its information coverage, while actively exploring the application of this data science technology to statistical work across more diverse industries, empowering official statistical services through innovative technology.