The futility of STILTs for the classification of lexical borrowings in Spanish
This work addresses the problem of automatic detection of lexical borrowings in Spanish for NLP researchers, showing incremental results by comparing methods without achieving new state-of-the-art performance.
The study tested supplementary training on intermediate labeled-data tasks (STILTs) for classifying lexical borrowings in Spanish, finding that STILTs provided no improvement over direct fine-tuning of multilingual models, with multilingual models on small language subsets performing better than multilingual BERT but worse than multilingual RoBERTa.
The first edition of the IberLEF 2021 shared task on automatic detection of borrowings (ADoBo) focused on detecting lexical borrowings that appeared in the Spanish press and that have recently been imported into the Spanish language. In this work, we tested supplementary training on intermediate labeled-data tasks (STILTs) from part of speech (POS), named entity recognition (NER), code-switching, and language identification approaches to the classification of borrowings at the token level using existing pre-trained transformer-based language models. Our extensive experimental results suggest that STILTs do not provide any improvement over direct fine-tuning of multilingual models. However, multilingual models trained on small subsets of languages perform reasonably better than multilingual BERT but not as good as multilingual RoBERTa for the given dataset.