AI
HIPE-2026 Benchmark Targets Accurate Person-Place Relation Extraction in Historical Texts
A new CLEF evaluation lab challenges systems to extract multilingual person-place associations from noisy historical documents while balancing accuracy, efficiency, and generalization.
Researchers have released HIPE-2026, the latest iteration of a CLEF evaluation lab designed to advance person-place relation extraction from multilingual historical texts. Building on prior HIPE campaigns (2020, 2022), this benchmark extends toward semantic relation extraction tasks in digital humanities and knowledge-graph construction.
The Task#
The evaluation lab asks systems to identify and classify two types of person-place relations in historical documents across multiple languages and time periods:
- at: Has the person ever been at this place?
- isAt: Is the person located at this place around publication time?
Both relation types require systems to reason over temporal and geographical cues embedded in noisy, often poorly preserved historical sources.
Evaluation Framework#
HIPE-2026 introduces a three-fold evaluation profile that jointly measures:
- Accuracy — How well systems extract and classify relations
- Computational efficiency — Resource consumption and inference speed
- Domain generalization — Performance across different language pairs and historical periods
Applications#
The benchmark targets downstream use cases in digital humanities and historical research:
- Knowledge-graph construction from archives
- Historical biography reconstruction
- Spatial analysis and historical geography
The lab reflects growing interest in applying NLP techniques to large-scale digitized historical collections, where noisy OCR output and language variation pose significant challenges to information extraction systems.
