**Using AI to Assist in Extracting the Value of Online Opinions [Dr. Cheong's Talk Part 5]**
In the previous article, I mentioned that structuring big data lays the foundation for creating value from the data. How so? With the rapid development of ABC (Artificial Intelligence, Big Data, and Cloud Computing) in recent years, these three technologies have emerged and combined in ways that were previously thought impossible or not yet achievable. This convergence is creating immeasurable value and impact on businesses, governments, and various aspects of social life. According to McKinsey's latest report, "Artificial Intelligence: The Next Digital Frontier?", global tech giants invested over $30 billion in AI in 2016, with 90% spent on R&D and deployment and 10% on AI acquisitions. The report found that a few early adopters are already benefiting, creating pressure on other companies to catch up in digital transformation. AI applications are entering the final breakthrough stage.
While AI is advancing rapidly, I believe that aspects involving human cognition, contextual semantics, and emotional interpretation cannot be fully realized through AI alone at this stage. Recent popular applications, such as online opinion and social media monitoring and insights, differ from widely discussed AI applications like robots (e.g., AlphaGo), automated vehicles, behavioral perception, or computer vision. Online opinions, expressed through text, emojis, images, and videos on various online platforms like social media, forums, and online media, can be collected, structured, mined, and analyzed in real-time. The results provide governments, businesses, and organizations with insights to strategize or adjust policies, better responding to public feedback on government policies and consumer feedback on brand experiences and services.
Four Key Charts May Not Be Accurate Enough for Decision Making
More than a decade ago, when online opinion analysis began to develop, it was mainly technology-driven. Industry professionals focused on data structuring and matching online text with positive or negative sentiments using linguistic methods. The analysis results were primarily presented in charts, commonly known as the "Four Key Charts": distribution of opinion sources, trend changes of a specific topic, sentiment distribution, and word clouds indicating popular concerns. While these charts allowed decision-makers to understand the distribution and focus of online opinions, the accuracy of sentiment analysis was low and ambiguous, and the charts provided limited insight and decision-making information.
In recent years, with the leap in machine learning development, AI assistance combined with manual coding has added a deeper layer of analysis beyond the Four Key Charts. Using big data technology and machine learning, unknown online opinion patterns are first identified from vast amounts of data. A professional team with industry knowledge then sets the direction and categories for in-depth analysis within these identified patterns, verifying data involving human cognition, contextual semantics, and emotions. This human-machine combination can address the nuances of online opinions, often filled with slang, irony, contextual logic, and conditional intentions. For example, a Facebook comment saying, "What a surprise, today's queue for getting a certificate was so fast, no friends!" clearly expresses satisfaction and a positive sentiment, but a machine might judge it as negative. Similarly, a hotel review saying, "The room is quite large and conveniently located, but the front desk staff were impolite," might be interpreted by a machine as mostly positive, but for the hotel, it's a negative review because room size and location can't be changed, and relying solely on machine judgment would miss the opportunity to improve front desk service.
Today, online opinions have become barometers influencing business development and government policies. AI assistance combined with professional human judgment significantly enhances the ability to process vast amounts of data and extract valuable insights.
Dr. Angus Cheong Chairman of the Asia-Pacific Internet Research Alliance and Chief Data Consultant at uMax Data Technology Ltd.
(Originally published in Hong Kong Economic Times, reprinted with permission)