Generative artificial intelligence-based modified abstractive cross attention enabled sequence to sequence model for abstractive Hindi text summarization

Abstract

Abstractive Text summarization is the process of providing a concise as well as cohesive summary, which encapsulates the vital information from the original text. Although the supervised models now in use are competent, they frequently rely on annotated datasets and create challenges regarding uninterpretability and limited generalization ability. To overcome the limitations, this research proposes a Generative Artificial Intelligence Sequence to sequence the Bidirectional Encoder Representations from the Transformers model to generate concise summaries. During the generation of the output sequence, the model takes into account data from various segments of the input sequence by utilizing a modified abstractive cross-attention technique. Specifically, the Generative Artificial Intelligence assists in removing grammatical mistakes in summaries via the application of the Global Surrogate method, which ensures the clarity and fluency of the output summary. In addition, the encoder-decoder model enables the accurate summary generation process, vastly improving fluency and accuracy. Furthermore, the experimental outcomes show that the Generative Artificial Intelligence Sequence to sequence the Bidirectional Encoder Representations from the Transformers model surpasses the conventional text summarization techniques concerning Bilingual Evaluation Understudy of 0.71 and Metric for Evaluation of Translation with Explicit Ordering of 0.73, which shows that the proposed model generates a meaningful summary from the given text.