Beyond Embeddings: Comparing Language-Aware Sparse Representations and Fine-Tuned Multilingual Transformer Classifiers for Multi-label Subject Classification in Institutional Repositories
| cic.institucionOrigen | Centro de Servicios en Gestión de Información | |
| cic.isFulltext | SI | |
| cic.isPeerReviewed | SI | |
| cic.lugarDesarrollo | Proyecto de Enlace de Bibliotecas | |
| cic.lugarDesarrollo | Servicio de Difusión de la Creación Intelectual | |
| cic.parentType | Objeto de conferencia | |
| cic.version | Publicada | |
| dc.date.accessioned | 2026-10-27T10:38:02Z | |
| dc.date.available | 2026-10-27T10:38:02Z | |
| dc.identifier.uri | https://digital.cic.gba.gob.ar/handle/11746/12824 | |
| dc.title | Beyond Embeddings: Comparing Language-Aware Sparse Representations and Fine-Tuned Multilingual Transformer Classifiers for Multi-label Subject Classification in Institutional Repositories | es |
| dc.type | Artículo | |
| dcterms.abstract | Institutional repositories need scalable support for multilingual subject metadata assignment to long documents. On 19,710 SEDICI items and 37 labels, we compare 231 sparse/fixed-embedding configurations and two fine-tuned DistilUSE/LaBSE classifiers under separate threshold-selection, validation, and test partitions. Language-aware BM25 full text with LinearSVC achieved held-out Macro-F1 = 0.7522 (95% CI 0.7372–0.7643), Micro-F1 = 0.7685 (0.7567–0.7797), and Recall@3 = 0.9320; the LaBSE-initialized classifier reached Macro-F1 = 0.7069. A post-hoc BM25–LaBSE matched-exposure control, validation-only over three seeds with no test access, preserved the aggregate BM25 advantage. Strong sparse baselines therefore remain competitive under the evaluated Transformer protocols. | en |
| dcterms.creator.author | Nusch, Carlos Javier | |
| dcterms.creator.author | Cagnina, Leticia Cecilia | |
| dcterms.creator.author | Lira, Ariel Jorge | |
| dcterms.creator.author | Villarreal, Gonzalo Luján | |
| dcterms.creator.author | Antonelli, Rubén Leandro | |
| dcterms.creator.author | Errecalde, Marcelo Luis | |
| dcterms.creator.author | Bértoli, Rafael | |
| dcterms.creator.author | Rojas, Manuel | |
| dcterms.creator.author | Vila, María Marta | |
| dcterms.creator.author | De Giusti, Marisa Raquel | |
| dcterms.extent | 811-820 | |
| dcterms.isPartOf.issue | XXXII Congreso Argentino de Ciencias de la Computación (Concepción del Uruguay, 5 al 9 de octubre de 2025) | |
| dcterms.isPartOf.item | Libro de actas - XXXII Congreso Argentino de Ciencias de la Computación - CACIC 2026 | |
| dcterms.language | Inglés | |
| dcterms.license | Attribution-NonCommercial-ShareAlike 4.0 International (BY-NC-SA 4.0) | |
| dcterms.subject | multi-label classification | es |
| dcterms.subject | automated subject indexing | es |
| dcterms.subject | institutional repositories | es |
| dcterms.subject | multilingual preprocessing | es |
| dcterms.subject | BM25 | es |
| dcterms.subject | DistilUSE | es |
| dcterms.subject | LaBSE | es |
| dcterms.subject.materia | Ciencias de la Computación |
Archivos
Bloque original
1 - 1 de 1
Cargando...
- Nombre:
- Nusch et al. - Beyond Embeddings - actas CACIC 2026-.pdf-PDFA.pdf
- Tamaño:
- 667.27 KB
- Formato:
- Adobe Portable Document Format
Bloque de licencias
1 - 1 de 1
Cargando...
- Nombre:
- license.txt
- Tamaño:
- 3.46 KB
- Formato:
- Item-specific license agreed upon to submission
- Descripción: