Post by ummefatihaayat22 on Feb 13, 2024 5:06:08 GMT -5
Natural Language Processing (NLP) is one of the Artificial Intelligence techniques that currently arouses the most interest among the research community and industry. This is why, in addition to events and conferences, it is also the protagonist of more and more competitions or challenges to test the NLP and advance accordingly with new solutions. IberLEF is one of those meetings, focused on NLP tasks in Spanish and with great international participation. Researchers have between 1 and 3 months to face the proposed challenges and then the best results are presented within the framework of the annual conference of the Spanish Society for Natural Language Processing (SEPLN) , which this year was held in A Coruña.
Alejandro Vaca, data scientist from the Institute of Germany Telemarketing Data Knowledge Engineering (IIC), was there, who participated in three of the challenges of this edition (LivingNER, QuALES and EXIST 2022) and managed to present his results. The papers of these investigations in PLN are now available. Detection of mentions of living beings One of IberLEF's challenges was to develop a solution to detect and link in texts any mentions of living beings that may exist LivingNER.
Clear-communicationIn the article “ Named Entity Recognition For Humans and Species With Domain-Specific and Domain-Adapted Transformer Models , Alejandro Vaca explains how he adapted a general NLP model to the biomedical domain , with the aim of improving the subsequent adjustment of the model to this task. . Indeed, the results show the effectiveness of using models adapted to a specific linguistic domain.
Additionally, it investigated the impact that the base vocabulary of a model has on tasks in a specific domain, showing the advantage that models trained specifically for a domain have in entity detection tasks . He finally obtained the second best result in the main task of the competition: entity detection.
Learn to answer questions in Spanish In the Learning Answers to Questions from Examples in Spanish (QuALES) challenge, Alejandro Vaca tested three Spanish language models , such as RigoBERTa . He trained and retrained them with various datasets and added the predictions to get more reliable answers to the questions. He explains the process and results in the article “Adversarial Question Answering in Spanish with Transformer Models.