ABSTRACTIVE SUMMARIZATION USING CATEGORICAL GRAPH NETWORK
The rapid development of technologies produce enormous amount of data which have lot of hidden insights. Extracting these hidden insights are challengeable for researchers and industrialists. Most of the data are in textual and unstructured format. Text mining is the prominent research area that has being utilized for the textual data analysis. Document summarization is an effective application which provides the summary of given content. This research work mainly focused on generating abstractive summarization from the multiple documents. It contributes abstractive summarization using the categorical graph network. Lot of duplicate or redundant sentences are there in the multiple documents. Proposed CATSum, which is a graph based abstractive summarization technique that identifies the duplication based on the similarities of the sentences. The proposed technique used ALBERT encoder model to train the datasets. Then it has built the content summary based on the connection between the sentences. The proposed work is measured using the ROUGE-1, ROUGE-2 and ROUGE-L metrics and produced better accuracy than the baseline methods.
Abstractive Text Summarization, Categorical Graph Network, Multiple-document Summarization.