Document Summarization Using Multi-Objective Weighted Graph Embedding Feature and Optimized Rank Based Deep Classifier

The necessity for automatic text summarizing techniques is growing in the modern day. The primary goal of extractive multi-document summarizing techniques is to eliminate unnecessary information while simultaneously obtaining the essential content of a group of documents. Multi objective weighted graph embedding feature with an Optimized rank based deep Convolutional Neural Network-Bidirectional Long Short Term Memory classifier (MOWGEF+rank deep CNN-BiLSTM model) is developed by this study for multi-document text summarization. The process of text summarization begins with two stages preprocess document to prepare for input as a multi-document summary and extract multi-objective weighted graph embedding features from its content. The profound achievement of this paper is to regard weighing model as a weighting for semantic information of word, phrase and sentence relationships. This model also has a conventional entropy-based feature selection technique that aids in finding significant variables. For the optimized rank-based deep classifier model these attributes are utilized as the input which captures the document intricate details. The optimized rank-based deep classifier's novel layer assesses words according to the significance they have become in a document. This arranging takes place using a novel optimization method that generates sentence ranking, demonstrating a unique and effective approach to multi-document summarizing. Elytron Vespid-B optimization (EVO) algorithm essentially takes inspiration from the foraging behaviors of insects, fine-tune the classifier's parameters to optimize its performance. Based on the achievements the model attains high recall 1, recall 2 and recall L values of 0.77, 0.39 and 0.75 respectively for DUC 2002 database.