IMPROVING COVERAGE AND NOVELTY OF ABSTRACTIVE TEXT SUMMARIZATION USING TRANSFER LEARNING AND DIVIDE AND CONQUER APPROACHES

Automatic Text Summarization (ATS) models yield outcomes with insufficient coverage of crucial details and poor degrees novelty. The first issue resulted from the lengthy input, while second problem characteristics training dataset itself. This research employs divide-and-conquer approach to address by breaking input into smaller pieces be summarized, followed conquest results in order cover more significant details. For challenge, these chunks are summarized trained on datasets higher novelty levels produce human-like concise summaries novel words that do not appear article. demonstrate an improvement both levels. Moreover, we defined a new metric measure summary. Finally, investigated findings discover whether is influenced itself, as CNN/DM, or model its objective, Pegasus.