Chinese Text Summarization Based on Multi-Layer Attention

Jiecheng Jiang

2024

Abstract

The increasing volume of textual data on the internet leads individuals to spend more time sifting through and identifying crucial information within texts. Automatic summarization technology emerges as a method to extract key information from lengthy texts, reducing the time required for information retrieval in the age of information overload, thus garnering increased attention from researchers. Automatic summarization technology can be categorized into extractive summarization, which relies solely on the original text content and has its limitations, and generative summarization, which offers greater flexibility. However, challenges persist in maintaining sufficient information integrity during text initialization and ensuring the generation of high-quality summaries in Chinese. This paper proposes a Multi-Layer attention model to solve Chinese text summarization. This model obtains a 39.51 ROUGE-1 score and 37.25 ROUGE-L on the LCSTS validation dataset. In addition, a model acceleration method is proposed, which uses 1x1 convolution kernel to replace the linear layer in encoder-decoder to reduce size of the neural network and improve speed of summary generation.

Paper Citation

in Harvard Style

Jiang J. (2024). Chinese Text Summarization Based on Multi-Layer Attention. In Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE; ISBN 978-989-758-690-3, SciTePress, pages 565-568. DOI: 10.5220/0012839000004547

in Bibtex Style

@conference{icdse24,
author={Jiecheng Jiang},
title={Chinese Text Summarization Based on Multi-Layer Attention},
booktitle={Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE},
year={2024},
pages={565-568},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012839000004547},
isbn={978-989-758-690-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE
TI - Chinese Text Summarization Based on Multi-Layer Attention
SN - 978-989-758-690-3
AU - Jiang J.
PY - 2024
SP - 565
EP - 568
DO - 10.5220/0012839000004547
PB - SciTePress