Controlling Length in Abstractive Summarization Using a Convolutional Neural Network

By Yizhu Liu, Zhiyi Luo, Kenny Q. Zhu. 2018

81 citations recorded by the source corpus.

Convolutional neural networks (CNNs) have met great success in abstractive summarization, but they cannot effectively generate summaries of desired lengths. Because generated summaries are used in difference scenarios which may have space or length constraints, the ability to control the summary length in abstractive summarization is an important problem. In this paper, we propose an approach to constrain the summary length by extending a convolutional sequence to sequence model. The results show that this approach generates high-quality summaries with user defined length, and outperforms the baselines consistently in terms of ROUGE score, length variations and semantic similarity.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

fairseq: A Fast, Extensible Toolkit for Sequence Modeling (2019, background): “…2018; Wu et al., 2019), language modeling (Dauphin et al., 2017; Baevski and Auli, 2019), abstractive document summarization (Fan et al., 2018a; Liu et al., 2018; Narayan et al., 2018), story generation (Fan et al., 2018b, 2019), error correction (Chollampatt and Ng, 2018), multilingual…”
CTRLsum: Towards Generic Controllable Text Summarization (2020, background): “In contrast, typical controllable summarization meth-ods (Fan et al., 2018; Liu et al., 2018; Tan et al., 2020; Suhara et al., 2020; Chan et al., 2021) prede-fine the specific control aspect (e.g. entity, length, topic, etc.) and rely on the corresponding control annotations during training.”
GSum: A General Framework for Guided Neural Abstractive Summarization (2020, background): “There are also works on controlling the summary length (Kikuchi et al., 2016; Liu et al., 2018b) and styles (Fan et al., 2018) by explicitly feeding the desired features to the model.”
How to Compare Summarizers without Target Length? Pitfalls, Solutions and Re-Examination of the Neural Summarization Literature (2019, background): “There are some exceptions (Ma and Nakagawa, 2013; Kikuchi et al., 2016; Liu et al., 2018), but starting with (Rush et al., 2015), systems produce summaries of variable length.”
Neural Machine Translation: A Review (2019, unknown): “decoder network [313, 314], at each time step as the number of remaining tokens [315], or by modifying Transformer positional embeddings [316].”
Addressing Semantic Drift in Generative Question Answering with Auxiliary Extraction (2021, background): “Recently, some studies have been carried out on increasing the reliability of generation in the encoder-decoder framework(Liu et al., 2018; Li et al., 2018).”