Enhancing Argumentative Dialogue Systems with RoBERTa and BiLSTM for Opinion Building in Natural Language Understanding

Natural Language Understanding(NLU) plays an main role in conversational AI, enables the machines to respond for human input and present the output utterance. This paper presents a Natural Language Framework that combines RoBERTa for embedding extraction and BiLSTM for context sequence modelling. This model performs intent classification on user utterance and semantic similarity for argumentative template selection. The experimental results represents the effectiveness of this approach achieving accuracy of 94.2 % and Fl score of 93.8% on Banking77 dataset and STS benchmark dataset. And also it out perform BERT and LSTM base models.The proposed architecture is flexible for real world applications, such as virtual assistants and chatbots, by employing pre trained transformer models and sequential learning techniques.