COMPARATIVE ANALYSIS OF THE THREE STATE-OF-THE-ART TRANSFORMER-BASED SEQ2SEQ ABSTRACTIVE SUMMARIZATION MODELS
Loading...
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
2021 International Young Scholars' Conference
Abstract
Abstract Today everyone can clearly see that amount of information on the internet growth rapidly. Because of it, people meet with huge problem - to process and get the most important parts from them. Thus, there is become a need on its clear and cost-effective summarization. Main goal of text summarization is to generate concise and accurate summary from input documents. In recent years, text summarization become one of the most topical subject in tech sphere. Big Tech companies like Facebook, Google, and Microsoft understood the importance of automatic summarization technologies, and not long ago published their results called BART, PEGASUS and ProphetNet respectively. This works showed the best results on various datasets with different sizes. The main idea behind this paper is to compare and analyze those models in terms of speed, accuracy, accessibility and other characteristics. In the next work, the results of the research are planned to be applied on Russian and Kazakh language datasets.
Description
Keywords
summarization, BART, PEGASUS, ProphetNet, 2021 International Young Scholars' Conference, №10
Citation
Ospan Smagul / COMPARATIVE ANALYSIS OF THE THREE STATE-OF-THE-ART TRANSFORMER-BASED SEQ2SEQ ABSTRACTIVE SUMMARIZATION MODELS / 2021 International Young Scholars' Conference