Investigating the use of deep learning for textual entailment in BRACU-NLI dataset

Abstract

This work aims to analyze the potential of deep neural models for text-based entailment in Bangla Language. Entailment is the method of determining whether one text infers or goes against another text. The study concentrates on the application of deep learning methods, such as Recurrent Neural Networks (RNNs), BERT, GPT for solving text-based entailment. The neural network method is trained to foretell the relationship between two text sequences, such as whether one text sequence entails the other or whether one text sequence provides evidence for the other. Other tasks, such as question answering, can also be tackled by fine-tuning these models on specific datasets. The findings of this work will contribute to the development of further developed NLP systems that can perform complex reasoning and entailment tasks.