Researchers have developed a system to extract high-quality data from historical newspaper scans. The Institutional Newspapers Pipeline uses a multi-step process to segment and analyze the text in each scan, including optical character recognition (OCR), type classification, and named entity recognition. This has resulted in a dataset of over 16 billion tokens extracted from 1.4 million public domain newspaper scans published between 1795 and 1930. The system is designed to be interpretable and customizable, and the authors have released their methods, models, and dataset as an open resource. --- Why it matters: This work matters because it provides a scalable solution for accessing large amounts of historical text data, which can be used for training machine learning models or conducting research in fields such as natural language processing and information retrieval. The resulting dataset is also a valuable resource for researchers and historians alike. Source: https://arxiv.org/abs/2608.18972