Unlocking Textual Relationships: Deep Learning for Recognizing Entailment
Franco Antonio Pranata, Matthew Christopher Tjondropurnomo, Nicholas Lorenzo Setiawan, Muhammad Rizki Nur Majiid, Bambang Dwi Wijanarko · 2024
In the field of Natural Language Processing (NLP), understanding implicit meanings in texts is a complex and significant task. This study delves into the use of deep learning models to address these challenges, specifically by focusing on the implementation and evaluation of RoBERTa models. The research compares the performance of RoBERTa with other variations of BERT such as DistilBERT and DeBERTa using evaluation metrics such as accuracy, precision, recall, F1 score, BLEU score and Cohen's Kappa. Findings revealed that RoBERTa outperformed other models achieving an accuracy rate of 88.05%, a precision level of 88.02 %, recall rate at 88.01 % and F1-score score of 88.01 %. Furthermore, with BLEU score of 0.783 and Cohen's Kappa at 0.778, RoBERTa demonstrates its ability to make predictions that are in line with expert annotations and reflect human judgments. This research highlights the potential for advanced BERT models to improve the accuracy and reliability of textual implication identification thus contributing towards more sophisticated and adaptable NLP applications.