Recent Advances in DL-based Text Summarization: A Systematic Review

The technique of creating a brief and relevant summary of a lengthy piece of text while retaining its vital content and overall relevance is known as text summarization. It involves condensing the original text while retaining its core message. In today's age of information overload, text summarization has gained immense significance as we encounter an excessive amount of textual information on a daily basis. Text summarization can be done manually by humans, but it can also be automated using ML techniques. DL models have demonstrated promising results in text summarization in recent years, and have become a major study area in the field of NLP. This study offers a synopsis of literature on the use of DL approaches for text summarization. The review covers various techniques such as CNN, RNN, LSTM, DeepSum, GA2C, Pointer Generator, and BERT, as well as various datasets such as CNN/Daily Mail and Arabic datasets. The ROUGE Score was used to assess the efficacy of text summarizing approaches, and BERT received the highest score of 98%. This study gives a detail examination of the current advantage in DL approaches for text summarization. Furthermore, it indicates possible new study areas in this subject.