A Proposed Methodology for Semantic Alignment and Specialization of Pre-trained Multilingual Embeddings Using Mixture-of-Experts and Contrastive Learning for Legal Text Retrieval in Ecuador
- 17 October 2025
- book chapter
- Published by Springer Nature
- p. 450-463
Abstract
No abstract availableKeywords
This publication has 11 references indexed in Scilit:
- Contrastive learning and mixture of experts enables precise vector embeddings in biological databasesScientific Reports, 2025
- Improving Alignment and Uniformity of Expert Representation with Contrastive Learning for Mixture-of-Experts Model2024
- UNSEE: Unsupervised Non-contrastive Sentence Embeddings2024
- M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation2024
- DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings2022
- Language-agnostic BERT Sentence Embedding2022
- SimCSE: Simple Contrastive Learning of Sentence Embeddings2021
- ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer2021
- Unsupervised Cross-lingual Representation Learning at Scale2020
- How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings2019