Abstract
Much of the data of interest does not reside in DBMSs, but in “almost” tabular files. The word “almost” hides many challenges: the data often lacks regularity, which prevents its automatic ingestion. In spreadsheet-style files, data cells may alternate in a more or less haphazard manner with label cells; some rows are aggregates calculated from others; a worksheet may contain multiple tables… Such files often contain statistical data, such as the production of various goods or birth rates by region. They are therefore invaluable for verifying the accuracy of claims made on these topics (“fact-checking”). We will present recent methods for finding and understanding these files, as well as for developing methods tailored to fact-checking.
