PROGRAMEXBR: A Dataset of Programming Exercises in Portuguese

Programming exercises are an important part of learning programming. There are tools, known as Online Judges, that provide resources for automatically assessing this type of activity against a series of pre-registered test cases. The data from these exercises can be used in various applications in the field of learning analytics. However, there is a lack of datasets with this availability for the Brazilian Portuguese language. The aim of this work is to collect and model a data repository for programming exercises, including instructions, test cases and exercise metadata. First, platforms are defined as data sources and Web Scraping techniques are applied to extract the desired content. From the collection, the data is modeled and made available in a public repository called PROGRAMEXBR. We hope that it can be useful in various contexts, be it for supporting teachers, training inference models and generating new exercises, or even for an application in learning analytics, such as recommendation methods in education.