Automatic Creation of Text Corpora for Low-Resource Languages from the Internet: The Case of Swiss German

TL;DR

This paper introduces SwissCrawl, a large Swiss German text corpus created via web scraping, demonstrating its usefulness for improving language modeling in low-resource languages.

Contribution

It presents a scalable method for building large text corpora from web data for low-resource languages, exemplified by Swiss German.

Findings

Significant improvement in language modeling performance using SwissCrawl

The web scraping approach is adaptable to other low-resource languages

Continuous updates enhance corpus relevance over time

Abstract

This paper presents SwissCrawl, the largest Swiss German text corpus to date. Composed of more than half a million sentences, it was generated using a customized web scraping tool that could be applied to other low-resource languages as well. The approach demonstrates how freely available web pages can be used to construct comprehensive text corpora, which are of fundamental importance for natural language processing. In an experimental evaluation, we show that using the new corpus leads to significant improvements for the task of language modeling. To capture new content, our approach will run continuously to keep increasing the corpus over time.