Identifying Rare Languages in Common Crawl Data is a Needles-in-a-Haystack Problem

Automatic language identification is frequently framed as a multi-class classification problem. However, when creating digital corpora, for less commonly written languages, it may be more appropriate to consider it a data mining problem. For these languages, one knows ahead of time that the vast majority of documents are of little interest. By minimizing resources spent on classifying such documents, we can create corpora for rare languages much faster and with better coverage than using established pipelines. To demonstrate the effectiveness of the language mining perspective, we introduce a new pipeline and corpora for several French-based Creoles.