@inproceedings{herman-etal-2026-feedfetcher,
    title = "{F}eed{F}etcher: A Resilient Web Feed Downloader for Corpus Construction",
    author = "Herman, Ond{\v{r}}ej  and
      Kraus, Jan  and
      Suchomel, Vit",
    editor = "Piperidis, Stelios  and
      Bel, N{\'u}ria  and
      van den Heuvel, Henk  and
      Ide, Nancy  and
      Krek, Simon  and
      Toral, Antonio",
    booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
    month = may,
    year = "2026",
    address = "Palma de Mallorca, Spain",
    publisher = "ELRA Language Resource Association",
    url = "https://preview.aclanthology.org/paragraph-normalization/2026.lrec-1.558/",
    doi = "10.63317/49txz3zreas2",
    pages = "7014--7022",
    abstract = "Building large-scale, timestamped monitor corpora requires robust and efficient tools for continuous web data acquisition. We present FeedFetcher, an open-source, lightweight yet resilient downloader designed to collect linguistic data from RSS/Atom web feeds. The tool enables continuous corpus updates by harvesting newly published web content with minimal downtime and high data integrity. Implemented in Rust for performance, memory safety, and scalable concurrency, FeedFetcher supports thousands of simultaneous connections while maintaining server politeness. The software is available under the GPL-3.0 license on \url{https://github.com/ondra/feed_fetcher}. In{~}our setup, the entire workflow integrates FeedFetcher with downstream text-processing pipelines for tokenization, lemmatization, corpus compilation and deployment. The system is currently used to update monitor corpora in 64 languages, producing approximately two billion tokens per month. These corpora are available in Sketch Engine. We also describe methods for discovering new web feeds, combining manual exploration with automated extraction from large-scale web crawls to expand linguistic coverage. We demonstrate the system{'}s applicability through a time-based analysis of word-frequency change, showing how long-term accumulation of timestamped data supports the study of lexical dynamics and language evolution."
}