Data-to-Text refers to the process of generating natural language descriptions from structured data (e.g., tables, graphs, query outputs) or unstructured data (e.g., images, video, audio). This capability is especially valuable in contexts where data needs to be explained, where large volumes of information exceed human processing capacity, or where users expect natural language responses—an increasingly common expectation in modern interfaces. This chapter begins by defining the Data-to-Text task and highlighting key use cases. It then explores the main challenges, focusing on two core subfields: Table-to-Text and Graph-to-Text and examining how earlier non Deep Learning approaches addressed them. The chapter concludes with a review of datasets and benchmarks relevant to Table-to-Text and Graph-to-Text tasks, providing a foundation for evaluating the solutions presented in the following chapter.