João Ruano | Searcharxiv

arXiv2025

Effective Multi-Task Learning for Biomedical Named Entity Recognition

Biomedical Named Entity Recognition presents significant challenges due to the complexity of biomedical terminology and inconsistencies in annotation across datasets. This paper introduces SRU-NER (Slot-based Recurrent Unit NER), a novel approach designed to handle nested named entities while integrating multiple datasets through an effective multi-task learning strategy. SRU-NER mitigates annotation gaps by dynamically adjusting loss computation to avoid penalizing predictions of entity types absent in a given dataset. Through extensive experiments, including a cross-corpus evaluation and human assessment of the model's predictions, SRU-NER achieves competitive performance in biomedical and general-domain NER tasks, while improving cross-domain generalization.

cs.CL↗

arXiv2024

Fourier-Mukai transform for fine compactified Prym varieties

Consider a finite covering $β: C \to X$ of a smooth projective curve $X$ by a reduced, projective, planar curve $C$. Associated to two general polarizations on $C$, $q$ and $q'$, one can construct the corresponding compactified Prym varieties $\overline{\mathrm{P}}_β(q)$ and $\overline{\mathrm{P}}_β(q')$. Consider $Γ$ to be the group of line bundles whose torsion coincides with the order of $β$. In this article we construct a Fourier-Mukai transform between the derived categories of $\overline{\mathrm{P}}_β(q)$ and the $Γ$-equivariant derived category of $\overline{\mathrm{P}}_β(q')$. Hence, we obtain a derived equivalence between the $\mathrm{SL}(n,\mathbb{C})$-Hitchin fibre and its associated $\mathrm{PGL}(n,\mathbb{C})$-Hitchin fibre for a dense class of singular spectral curves. Our work then provides the extension of the Fourier-Mukai transform constructed by Arinkin and Melo-Rapagnetta-Viviani, which corresponds to autoduality of $\mathrm{GL}(n,\mathbb{C})$-Hitchin fibres in this class of singular spectral curves.

math.AG↗