This paper introduces novel neural architectures for semantic role labeling in languages with limited training data, achieving state-of-the-art results across multiple language families.
A comprehensive review of contextual embedding methods, from early approaches to current transformer-based models, with empirical comparisons across multiple benchmarks.
We present a novel attention-based framework for cross-lingual document retrieval that outperforms traditional translation-based approaches.
Developing computational tools and resources for low-resource languages to promote linguistic diversity in AI applications.
Creating interpretable machine learning models for extracting clinical information from electronic health records while maintaining transparency for healthcare professionals.