Abstract: A method for providing a balanced training dataset for training a Machine Learning model includes receiving a multi-class dataset containing data records of at least one majority class and at least one minority class, representing the data records of the at least one majority class by using at least one content-based representation of the data records, k-means clustering of the data records of the at least one majority class based on the at least one content-based representation, selecting the data record closest to the centroid of each cluster as representative of the respective cluster, aggregating the selected data records of the at least one majority class and the data records of the at least one minority class, and providing the aggregated data records as training dataset. Also disclosed is a data processing device, a computer program product, a data carrier signal, and a method for selecting representatives of data records.

Abstract: The present invention enables the computation of various types of information for a particular scanned and OCR recognised or retyped historical input document. It provides a global view on the “patterns” for historical language variation (text profiling) and the OCR errors most frequently found in the text (error profiling). For each of the individual tokens of the OCR output, an interpretation is given which based on the document specific information attempts to describe both, the underlying correct word of the text and the corresponding modern spelling of the word. This not only provides input for optimised OCR recognition of historical documents, but also for quality assurance and improved information retrieval.