Abstract:Nowadays, due to the constantly growing amount of textual information, automatic text summarization constitutes an important research area in natural language processing. In this work, we present a novel framework that combines semantic graph representations along with deep learning predictions to generate abstractive summaries of single documents, in an effort to utilize a semantic representation of the unstructured textual content in a machine-readable, structured, and concise manner. The overall framework is based on a well defined methodology for performing semantic graph parsing, graph construction, graph transformations for machine learning models and deep learning predictions. The employed semantic graph representation focuses on using the model of abstract meaning representation. Several combinations of graph construction and graph transformation methods are investigated to specify the most efficient of them for the machine learning models. Additionally, a range of deep learning architectures is examined, including a sequence-to-sequence attentive network, reinforcement learning, transformer-based architectures, and pre-trained neural language models. In this direction, a semantic graph representation of an original text is extracted, and then the present framework formulates the problem as a graph-to-summary learning problem to predict a summary of an original text. To the best of our knowledge, this formulation of graph-to-summary predictions in abstractive text summarization, without other intermediate steps in the machine learning phase, has not been presented in the relevant literature. Another important contribution is an introduction of a measure for assessing the factual consistency of the generated summaries in an effort to provide a qualitative evaluation. To assess the framework, an extensive experimental procedure is presented that uses popular datasets to evaluate key aspects of the proposed approach. The obtained results exhibit promising performance, validating the robustness of the proposed framework.

Question‐driven text summarization using an extractive‐abstractive framework

Question-aware Transformer Models for Consumer Health Question Summarization

Unified extractive-abstractive summarization: a hybrid approach utilizing BERT and transformer models for enhanced document summarization

Reinforced Extractive Summarization with Question-Focused Rewards

Guiding Extractive Summarization with Question-Answering Rewards

Topic-Aware Abstractive Text Summarization

A Survey of the State-of-the-Art Models in Neural Abstractive Text Summarization

Combination of abstractive and extractive approaches for summarization of long scientific texts

Topic-Guided Abstractive Text Summarization: a Joint Learning Approach

Query-oriented unsupervised multi-document summarization via deep learning model

Abstractive Summarization Improved by WordNet-based Extractive Sentences

Reinforcement Learning for Abstractive Question Summarization with Question-aware Semantic Rewards

Abstractive text summarization: State of the art, challenges, and improvements

Abstractive Summarization Using Attentive Neural Techniques

Synthesizing Scientific Summaries: An Extractive and Abstractive Approach

Text summarization based on semantic graphs: an abstract meaning representation graph-to-text deep learning approach

Multi-Perspective Abstractive Answer Summarization

HITS-based attentional neural model for abstractive summarization

Abstractive summarization incorporating graph knowledge

Extractive Summarization As Text Matching

On Extractive and Abstractive Neural Document Summarization with Transformer Language Models