Cristina Garcia Holgado

Also published as: Cristina Garcia Holgado


pdf bib
Empowering Low-Resource Regional Languages with Lexicons : A Comparative Study of NLP Tools for Morphosyntactic Analysis
Cristina Garcia Holgado | Marianne Vergez-Couret
Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)

We investigate the effect of integrating lexicon information to an extremely low-resource language when annotated data is scarce for morpho-syntactic analysis. Obtaining such data and linguistic resources for these languages are usually constrained by a lack of human and financial resources making this task particularly challenging. In this paper, we describe the collection and leverage of a bilingual lexicon for Poitevin-Saintongeais, a regional language of France, to create augmented data through a neighbor-based distributional method. We assess this lexicon-driven approach in improving POS tagging while using different lexicon and augmented data sizes. To evaluate this strategy, we compare two distinct paradigms: neural networks, which typically require extensive data, and a conventional probabilistic approach, in which a lexicon is instrumental in its performance. Our findings reveal that the lexicon is a valuable asset for all models, but in particular for neural, demonstrating an enhanced generalization across diverse classes without requiring an extensive lexicon size.


pdf bib
IAI @ SocialDisNER : Catch me if you can! Capturing complex disease mentions in tweets
Aman Sinha | Cristina Garcia Holgado | Marianne Clausel | Matthieu Constant
Proceedings of The Seventh Workshop on Social Media Mining for Health Applications, Workshop & Shared Task

Biomedical NER is an active research area today. Despite the availability of state-of-the-art models for standard NER tasks, their performance degrades on biomedical data due to OOV entities and the challenges encountered in specialized domains. We use Flair-NER framework to investigate the effectiveness of various contextual and static embeddings for NER on Spanish tweets, in particular, to capture complex disease mentions.