CzechTopic: A Benchmark for Zero-Shot Topic Localization in Historical Czech Documents
This work addresses the problem of topic localization in historical documents for historians and researchers working with Czech documents, providing an incremental step in the field of natural language processing.
The authors tackled the problem of zero-shot topic localization in historical Czech documents, achieving near-human topic detection with the strongest models, while others showed pronounced failures in span localization, with performance ranging from 0 to near-human agreement. The best models approached human agreement.
Topic localization aims to identify spans of text that express a given topic defined by a name and description. To study this task, we introduce a human-annotated benchmark based on Czech historical documents, containing human-defined topics together with manually annotated spans and supporting evaluation at both document and word levels. Evaluation is performed relative to human agreement rather than a single reference annotation. We evaluate a diverse range of large language models alongside BERT-based models fine-tuned on a distilled development dataset. Results reveal substantial variability among LLMs, with performance ranging from near-human topic detection to pronounced failures in span localization. While the strongest models approach human agreement, the distilled token embedding models remain competitive despite their smaller scale. The dataset and evaluation framework are publicly available at: https://github.com/dcgm/czechtopic.