Article Preview
TopIntroduction
The world has been facing an unprecedented health crisis caused by the COVID-19. Government-coordinated efforts across the globe have focused on coping with the COVID-19 situation. Based on currently available information and clinical expertise, older adults might be at higher risk for severe illness from COVID-19 (Xie et al., 2020). As COVID-19 is a new virus, the public has lots of questions and confusion about this disease. Therefore, it is critical to provide timely, accurate, and relevant information to the public, which can be the difference between life and death. With lockdowns and other social distancing measures in effect in many countries, more and more people are relying on the Internet and digital tools for information and advice. Digital information solutions (e.g., (Chakraborty et al., 2020; Chakraborty & Rodrigues, 2020; Dash et al., 2021)) can be invaluable in the prevention and management of COVID-19. Over the past months, we have witnessed the increasing adoption of mobile apps to disseminate COVID-19 related information and resources. However, older adults’ adoption and use of digital technology lag that of younger cohorts. According to Norman & Skinner (Norman & Skinner, 2006), older adults’ eHealth literacy, or the ability to access, assess, and use health information to make informed healthcare decisions tends to be low and requires extensive assistance. Compared with younger people, older adults are less likely to obtain high-quality information or obtain food, supplies, and services online (Xie et al., 2020). The complexity of these digital tools makes them intimidating for older adults who struggle to use a computer or smartphone due to inexperience with the computer.
In this paper, we present the design of a chatbot assistant that offers real-time, useful, context-aware, and personalized tips and information to users, especially older adults through automatic question answering. The chatbot assistant provides a natural, interactive interface for users, and it does not require the user to have a technical background. To empower the chatbot with a “knowledgeable brain”, we construct a comprehensive knowledge graph that connects multiple knowledge sources related to COVID-19. A knowledge graph (KG) represents a collection of interlinked entities. The entities (nodes) in a KG represent real-world objects, while the links (edges) between two entities present the relationship between them. KG is vital in our system for information search, analytics, and recommendations. Moreover, we use KGs to represent our knowledge for the following reasons: (1) KGs have formal semantics that allows both people and computers to process them efficiently and unambiguously. (2) Questions/queries can be naturally mapped with KGs. (3) KGs can be easily expanded with new knowledge/data. Based on the knowledge graph, we build a flexible dialogue system to support task-oriented conversation that can recognize user’s speech, identify user’s intention, and provide suitable responses to users, using AI technologies including natural language processing, semantic search, logic reasoning, and machine learning. It can answer various questions about COVID-19, such as general questions like “Is COVID-19 vaccine safe?”, “What are the symptoms of COVID-19?”, “Is COVID-19 airborne?”, etc.; or specific personal questions, like “Do I need to be tested for COVID-19?”, “Where I can get tested for COVID-19?”, “How many new cases of COVID-19 appear in my grandson’s school this week?”, etc.
Overall, the contributions of the paper are summarized as follows:
• We identified people’s information need during the COVID -19 pandemic and identified and prepared comprehensive information datasets that have been used as knowledge sources to get people informed of this disease.
• We developed a chatbot prototype system using AI technologies including natural language processing, knowledge graph, logic reasoning, and machine learning. This chatbot aims to assist people, especially older adults, to get timely information about COVID-19. Design decisions were made considering older adults’ special needs.
• We evaluated the proposed chatbot with a set of qualitative evaluations.