CLEF25 LongEval: CIR_cluster at LongEval 2025: Clustering Query Variants for Temporal Generalization
We describe the participation of the CIR_cluster team in the CLEF 2025 LongEval WebSearch task. In longitudinal settings, approaches that leverage historical information—such as past relevance judgments—have demonstrated strong effectiveness. However, these methods are limited when no such information is available. For instance, relying on previous clicks is infeasible for queries that have never been issued before. We hypothesize that documents relevant to a given query are also relevant to its semantic variants. Based on this assumption, we cluster queries to identify query variants. This enables us to link previously unseen queries to the histories of its query variants. By that, the extended approaches can generalize not only to new and updated documents but also to new and updated queries. Our experimental evaluation showed that clustering did not improve the average retrieval effectiveness. However, when query variants could be identified, the performance often polarizes—resulting in either substantial improvements or declines. Although our current approach did not yield overall performance improvements, we think that identifying query variants remains an interesting direction to generalization across queries for ranking approaches that employ prior relevance signals in longitudinal settings.
@inproceedings{Ndiema2025,
author = {Arthur Muanza Ndiema AND J{\"{u}}ri Keller and Philipp Schaer},
title = {{LongEval: CIR\_cluster at LongEval 2025: Clustering Query Variants for Temporal Generalization}},
booktitle = {Working Notes of CLEF 2025 -- Conference and Labs of the Evaluation Forum, CEUR Workshop Proceedings},
editor = {Faggioli, Guglielmo and Ferro, Nicola and Rosso, Paolo and Spina, Damiano},
year = {2025}
}