A Hierarchical Model for Text Autosummarization

The mxnet implementation of the project of A Hierarchical Model for Text Autosummarization

Abstract

Summarization is an important challange in natural language processing. Deep learning methods, however, have not been widely used in text summarization, although neural networks have been proved to be powerful in natural language processing. In this paper, an encoder-decoder neural network model is applied to text summarization, as an important step toward this task. Besides, a hierarchical model, which builds the sentence representations and then paragraph representations, enables the summarization for long documents.

Usage

run

python auto_sum_lstm.py

to train the model

run

python validation.py

to evaluate the model on validation set