Orientador(es)
Resumo(s)
We describe IPR's systems developed for ASSIN2 (Evaluating Semantic Similarity and Textual Entailment). Our best submission ranked first in the Semantic Textual Similarity task and second in the Recognizing Textual Entailment task. These systems were developed using BERT, for each task we added one layer to a pre-trained Bert model and fine-tuned the whole task network.
Descrição
Palavras-chave
General Computer Science
