The World Wide Web is a major source of textual information, with a human-readable semi-structured format, referring to multiple domains, some of them highly complex. Traditional ETL approaches following the development of specific source code for each data source and based on multiple domain / computerscience experts interactions, become an inadequate solution, time consuming and prone to error. This paper presents a novel approach to ETL, based on its decomposition in two phases: ETD (Extraction, Transformation and Data Delivery) and IL (Integration and Loading). The ETD proposal is supported by a declarative language for expressing ETD statements and a graphical application for interacting with the domain expert. When applying ETD mainly domain expertise is required, while computer-science expertise will be centred in the IL phase, linking the processed data to target system models, enabling a clearer separation of concerns. This paper presents how ETD has been integrated, tested and validated in a space domain project, currently operational at the European Space Agency for the Galileo Mission.