Call For Participation HIPE 2026 – CLEF Shared Task on Person-Place Relation Extraction from Multilingual Historical Texts
*(apologies for cross-postings)* ------------------------------
*HIPE:* Identifying Historical People, Places and other Entities. *Website:* https://hipe-eval.github.io/HIPE-2026/ *Tasks:* Person-Location Relation Extraction from Multilingual Historical Texts. *Registration:* https://clef-labs-registration.dipintra.it/ (until 23 April 2026) *Training data releases:* 19 Dec 2025 (partial); 19 Jan 2026 (full) *Evaluation period:* 5–7 May 2026 *Workshop venue:* during CLEF conference, 21–24 September 2026, Jena, Germany. *LinkedIn:* @ImpressoProject / #HIPE2026 / @clef_initiative / #clef2026 ------------------------------ "Who was where when?"
We invite participation in the third edition of the HIPE shared task, dedicated to the extraction of person–place relations in multilingual historical documents. Building on the success of HIPE-2020 and HIPE-2022, which focused on entity recognition and linking, HIPE-2026 aims to enable finer-grained analysis of entities and support the accurate reconstruction of individuals’ geographical and temporal trajectories.
The objective of HIPE-2026 is to *build systems capable of determining whether a relation holds between a person and a location (place) mentioned in a document*, and classify its temporal scope. Participants are asked to develop systems that determine, for each (person, location) pair associated with a historical document, whether the text implies that the person is at that location within the document’s temporal horizon (isAt relation), or that the person was there at some earlier moment in their life (a more general At relation), or that no such link can be established.
Can large language models take up the challenges? Simple co-occurrences of entity mentions in a text are not sufficient to uncover the implicit and explicit, temporally anchored relations between person and locations. Addressing this challenge requires temporal reasoning, geographical inference, and the interpretation of noisy historical texts (often with only fragmentary contextual cues) to classify person–location relations with varying degrees of certainty.
The task is designed to be tackled by generative AI systems/LLMs as well as by more traditional classification approaches. HIPE-2026 features two evaluation profiles
- *Accuracy Profile*: Focusing on system performance in relation classification. - *Efficiency Profile*: Rewarding scalable, lightweight approaches considering model size and compute cost. - *Generalization Profile*: An unseen dataset from a different domain will be included to evaluate systems’ ability to generalise beyond the newspaper domain data.
For the accuracy and efficiency profile, training and test data originate from historical newspapers in English, German, French and Luxembourgish.
Entity pairs will be provided. For further information on data, tasks, and evaluation settings
- HIPE-2026 website: https://hipe-eval.github.io/HIPE-2026/ - Participation Guidelines: https://doi.org/10.5281/zenodo.17800136 - HIPE-2026-data GitHub repository: https://github.com/hipe-eval/HIPE-2026-data
On HIPE shared tasks
HIPE evaluation lab series is part of the ongoing efforts of the natural language processing and digital humanities communities to adapt and develop technologies to efficiently retrieving and exploring information from historical texts. Important dates
- 17 Nov 2025: Lab registration opens. - 03 Dec 2025: Release of example data. - 19 Dec 2025: Release of partial training data. - 19 Jan 2026: Release of final training data. - 23 Apr 2026: Lab registration closes. - 05 May 2026: Test data release (10:00 CEST). - 07 May 2026: Participant run submission deadline. - 13 May 2026: Publication of results and release of test data. - 28 May 2026: Submission of participant notebook paper. - 10 Jul 2026 / 31 Aug 2026: CLEF conference regular/late registration DL. - 21 Sep 2026: CLEF 2026 Conference.
Best regards, *HIPE-2026 Shared Task Organizers* https://hipe-eval.github.io/HIPE-2026/