Automatic quiz generation for elderly people
Abstract
Studies have indicated that games can be beneficial for the elderly, in areas such as cognitive functioning and well-being. Taking part in social activities, such as playing a game with others, could also be beneficial. One type of game is a computer-based quiz. One can create quiz questions manually; however, this can be time-consuming. Another approach is to generate quiz questions automatically. This project has examined how quizzes for Norwegian elderly can be automatically generated using online resources. To get a better understanding of the topic, a multiple choice quiz application (prototype) has been developed, which automatically generates questions and answers. The target users for the quiz are elderly people above the age of 60. The quiz can ask factual questions about people (actors and musicians) that were well-known and works (movies and literature) that were created during its target users' past. The quiz can be played alone or with another person. Creating the quiz involved retrieving web pages from Wikipedia, using information extraction to extract quiz data from those web pages, and representing and storing the data with Semantic Web technologies. Some of the quiz data was obtained from DBpedia, which makes available (mostly) structured data from Wikipedia for storage. In addition, two approaches were used in this project for identification or extraction of data, which could then be represented and stored to support or enable further question generation. The first approach is to create a Document Object Model tree from relevant web pages, and use extraction rules to extract data from semi-structured HTML elements (lists and tables). The second approach is to detect and classify entities and their relationships in unstructured text. All quiz data was extracted from Wikipedia, but the approaches used by the quiz application and DBpedia all target information from different parts of Wikipedia articles. The quiz generated in this project is in Norwegian, but methods used are generic and it is easy to generate quizzes in other languages. The quiz was evaluated by means of usability testing with participants, to get an indication of its usability according to specific usability measures. The majority of the participants thought the quiz was useful and effective to use, and expressed satisfaction with the quiz. Some potential improvements were also identified during the evaluation.