Data-Driven Assistance for User Decision Making on Mobile Devices.

다국어 초록 (Multilingual Abstract)

Mobile devices are ubiquitous. As of 2019, two-thirds of the world population own a mobile phone. Mobile devices are indispensable for supporting billions of users' information access activities such as searching, browsing news, and shopping. Among those activities, users may often need to make decisions when the mobile device is the only available channel for their information access. However, users' mobile decision-making experience is hindered by the physical characteristics of mobile devices: they are small and it is difficult to type on these devices. Furthermore, both editing and navigation would be harder than that on computers. These characteristics result in more difficulties for users to search, digest and compare information, which are the necessary steps in the process of decision making. Can we make it very easy for users to make decisions on mobile devices? In this dissertation, for the first time, we investigate the techniques for improving users' mobile decision making experience as a whole. We identify that the key to assisting user decision making is through suggesting external knowledge to bridge their knowledge gap. To this end, we propose to learn or mine such external knowledge from massive mobile-related data. We investigate three important real-world decision-making problems on mobile devices: mobile shopping decisions (Chapter 2), security decisions (Chapter 2 and Chapter 4), and business decisions (Chapter 5). We bridge users' knowledge gap in the following ways. In the first problem, we leverage a search-engine log to expand the missing information in user queries (Chapter 2); in the second problem, we leverage the Google Playstore meta-data to retrieve explanatory information to directly address users' confusion (Chapter 3 and Chapter 4), finally, in the third problem, we leverage text-to-SQL data to generate SQL from a natural language question, so that users can easily query the database using natural language (Chapter 5). Our experimental results prove that massive mobile-related data can be leveraged to effectively assist users' mobile decision making by suggesting external knowledge.