Fractional Snow Ablation Optimization-based Random Multimodel Deep Learning for Web Document Summarization with Large Language Model

The substantial amount of information available on the web has made it challenging for users to search and process relevant content efficiently. Notably, web documents, such as product descriptions, blog posts, research papers, news stories, and articles, often contain lengthy and vast amounts of irrelevant or redundant information. Various methods have been introduced to summarize the web. However, they have faced challenges in mining key points from the documents. Hence, a new model termed Fractional Snow Ablation Optimization with Random Multimodel Deep Learning (FSAO-RMDL) is devised for web document summarization. Initially, the input web document undergoes tokenization, where the Bidirectional Encoder Representations from Transformers (BERT) is employed to tokenize the document. Following this, feature extraction is executed, and extractive summarization is performed using Random Multimodel Deep Learning (RMDL), trained by Fractional Snow Ablation Optimization (FSAO). The FSAO approach incorporates the Fractional Calculus (FC) and Snow Ablation optimization (SAO). Lastly, abstractive summarization is performed by exploiting the GPT-NeoX Large Language Model (LLM). Overall, the experimental outcomes of the FSAO_RMDL approach demonstrate that it obtained a maximum recall, F-measure, and precision of 93.766%, 92.750%, and 91.755%, respectively.