The DimABSA shared task aims to combine dimensional analysis with Aspect-Based Sentiment Analysis (ABSA). It addresses the lack of continuous sentiment representation, as opposed to categorical labels (e.g., positive, negative, or neutral), and enriches it with an assessment of arousal. Our team’s PUEB-DimASR investigates the "mean-regression trap" — the tendency of standard MSE loss in high-dimensional sentiment tasks to over-predict values closer to the global mean. We propose a two-step advancement in model ar chitecture. First, we enhance baseline Trans formers with Graph Convolutional Networks(GCN) to capture syntactic aspect-sentiment dependencies. Second, we evaluate and recommend a Hybrid loss function that combines Mean Squared Error (MSE) and Concordance Correlation Coefficient (CCC).Our proposed GCN-deBERTa model consistently outperforms the baseline across six target languages. While MSE loss yields the best RMSE scores for English (0.876) and Chinese (0.546), it introduces significant variance collapse, which we successfully mitigated using the Hybrid loss, achieving near-perfect distributional alignment (99.6\%). Additionally, our model trained with the Hybrid loss achieved the best RMSE scores for Russian (1.136), Tatar (1.207), and Ukrainian (1.178).

This paper reports on an endeavour of creating basic linguistic resources for geo-referencing of Polish free-text documents. We have defined a fine-grained named entity hierarchy, produced an exhaustive gazetteer, and developed named-entity grammars for Polish. Additionally, an annotated corpus for the cadastral domain was prepared for evaluation purposes. Our baseline approach to geo-referencing is based on application of aforementioned resources and a lightweight co-referencing technique which utilizes string-similarity metric of Jaro-Winkler. We carried out a detailed evaluation of detecting locations, organizations and persons, which revealed that best results are obtained via application of a combined grammar for all types. The application of lightweight co-referencing for organizations and persons improves recall but deteriorates precision, and no gain is observed for locations. The paper is accompanied by a demo, a geo-referencing application capable of: (a) finding documents and text fragments based on named entities and (b) populating the spatial ontology from texts.