AI Brown and AI Koditex: LLM-Generated Corpora Comparable to Traditional Corpora of English and Czech Texts

2025-09-26

AI Summary

A lack of comparable corpora hinders systematic linguistic analysis of differences between AI-generated and human-authored texts. Method: We constructed the first bilingual (English–Czech) LLM-generated corpus rigorously aligned with classical human-annotated reference corpora (Brown/Koditex) in genre, topic, and structural organization. Texts were generated using models from GPT-3 through GPT-4.5, then tokenized, lemmatized, and syntactically annotated following the Universal Dependencies framework, with standardized processing via multi-platform APIs. Contribution/Results: We release a 48.5-million-token open-source corpus (27M English, 21.5M Czech) under CC licensing, integrated into the Czech National Corpus. This resource enables, for the first time, fine-grained cross-lingual and cross-model comparative analysis of linguistic features, establishing a benchmark dataset for AI-text detection and studies of language evolution in generative AI contexts.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Crowd Sourcing and Human Computation

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

This article presents two corpora of English and Czech texts generated with large language models (LLMs). The motivation is to create a resource for comparing human-written texts with LLM-generated text linguistically. Emphasis was placed on ensuring these resources are multi-genre and rich in terms of topics, authors, and text types, while maintaining comparability with existing human-created corpora. These generated corpora replicate reference human corpora: BE21 by Paul Baker, which is a modern version of the original Brown Corpus, and Koditex corpus that also follows the Brown Corpus tradition but in Czech. The new corpora were generated using models from OpenAI, Anthropic, Alphabet, Meta, and DeepSeek, ranging from GPT-3 (davinci-002) to GPT-4.5, and are tagged according to the Universal Dependencies standard (i.e., they are tokenized, lemmatized, and morphologically and syntactically annotated). The subcorpus size varies according to the model used (the English part contains on average 864k tokens per model, 27M tokens altogether, the Czech partcontains on average 768k tokens per model, 21.5M tokens altogether). The corpora are freely available for download under the CC BY 4.0 license (the annotated data are under CC BY-NC-SA 4.0 licence) and are also accessible through the search interface of the Czech National Corpus.

Problem

• Creating LLM-generated corpora comparable to human-written English and Czech texts
• Developing multi-genre resources for linguistic comparison between human and AI writing
• Providing freely accessible annotated corpora replicating established human corpus structures

Innovation

• Generated multi-genre corpora using multiple LLMs
• Replicated human corpora structure with linguistic annotations
• Provided freely accessible resources under open licenses