Exploiting Discourse-Level Segmentation for Extractive Summarization

Liu, Zhengyuan; Chen, Nancy

doi:10.18653/v1/d19-5415

Cited by 23 publications

(19 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…C ¸elikyilmaz et al ( 2018) introduced a hierarchical model that handles the encoding phase through collaborating agents responsible for processing each text subsection. Liu and Chen (2019) and Xu et al (2020) proposed to exploit the discourse segmentation to extract the salient content for extractive summarization. Gidiotis and Tsoumakas (2020) introduced a divide-and-conquer approach that relies on structured documents to summarize each section independently.…”

Section: Related Workmentioning

confidence: 99%

Semantic Self-Segmentation for Abstractive Summarization of Long Documents in Low-Resource Regimes

Moro

Ragazzi

2022

AAAI

View full text Add to dashboard Cite

The quadratic memory complexity of transformers prevents long document summarization in low computational resource scenarios. State-of-the-art models need to apply input truncation, thus discarding and ignoring potential summary-relevant contents, leading to a performance drop. Furthermore, this loss is generally destructive for semantic text analytics in high-impact domains such as the legal one. In this paper, we propose a novel semantic self-segmentation (Se3) approach for long document summarization to address the critical problems of low-resource regimes, namely to process inputs longer than the GPU memory capacity and produce accurate summaries despite the availability of only a few dozens of training instances. Se3 segments a long input into semantically coherent chunks, allowing transformers to summarize very long documents without truncation by summarizing each chunk and concatenating the results. Experimental outcomes show the approach significantly improves the performance of abstractive summarization transformers, even with just a dozen of labeled data, achieving new state-of-the-art results on two legal datasets of different domains and contents. Finally, we report ablation studies to evaluate each contribution of the components of our method to the performance gain.

show abstract

Section: Related Workmentioning

confidence: 99%

Semantic Self-Segmentation for Abstractive Summarization of Long Documents in Low-Resource Regimes

Moro

Ragazzi

2022

AAAI

View full text Add to dashboard Cite

show abstract

“…In text summarization, most benchmark datasets focus on the news domain, such as NYT (Sandhaus, 2008) and CNN/Daily Mail (Hermann et al, 2015), where the human-written summaries are used in both abstractive and extractive paradigms (Gehrmann et al, 2018). To improve the performance of extractive summarization, non-neural approaches explore various linguistic and statistical features such as lexical characteristics (Kupiec et al, 1995), latent topic information (Ying-Lang Chang and Chien, 2009), discourse analysis (Hirao et al, 2015;Liu and Chen, 2019), and graphbased modeling (Erkan and Radev, 2004;Mihalcea and Tarau, 2004) . In contrast, neural approaches learn the features in a data-driven manner.…”

Section: In Relation To Other Workmentioning

confidence: 99%

Conditional Neural Generation using Sub-Aspect Functions for Extractive News Summarization

Liu

Shi

Chen

2020

Findings of the Association for Computational Linguistics: EMNLP 2020

Self Cite

View full text Add to dashboard Cite

Much progress has been made in text summarization, fueled by neural architectures using large-scale training corpora. However, in the news domain, neural models easily overfit by leveraging position-related features due to the prevalence of the inverted pyramid writing style. In addition, there is an unmet need to generate a variety of summaries for different users. In this paper, we propose a neural framework that can flexibly control summary generation by introducing a set of subaspect functions (i.e. importance, diversity, position). These sub-aspect functions are regulated by a set of control codes to decide which sub-aspect to focus on during summary generation. We demonstrate that extracted summaries with minimal position bias is comparable with those generated by standard models that take advantage of position preference. We also show that news summaries generated with a focus on diversity can be more preferred by human raters. These results suggest that a more flexible neural summarization framework providing more control options could be desirable in tailoring to different user preferences, which is useful since it is often impractical to articulate such preferences for different applications a priori.

show abstract

“…Several studies [35,37] set up document graphs according to discourse analysis. But, this method relies on external tools, which possibly bring out semantically fragmented outputs [20]. To sum up, GAT (Graph Attention Network) is constructed using sentence context representation and topic information, and the docu-ment context vector and topic information are updated simultaneously.…”

Section: Introductionmentioning

confidence: 99%

GATSum: Graph-Based Topic-Aware Abstract Text Summarization

Jiang¹,

Zou

Xu³

et al. 2022

ITC

View full text Add to dashboard Cite

The purpose of text summarization is to compress a text document into a summary containing key information. abstract approaches are challenging tasks, it is necessary to design a mechanism to effectively extract salient information from the source text, and then generate a summary. However, most of the existing abstract approaches are difficult to capture global semantics, ignoring the impact of global information on obtaining important content. To solve this problem, this paper proposes a Graph-Based Topic Aware abstract Text Summarization (GTASum) framework. Specifically, GTASum seamlessly incorporates a neural topic model to discover potential topic information, which can provide document-level features for generating summaries. In addition, the model integrates the graph neural network which can effectively capture the relationship between sentences through the document representation of graph structure, and simultaneously update the local and global information. The further discussion showed that latent topics can help the model capture salient content. We conducted experiments on two datasets, and the result shows that GTASum is superior to many extractive and abstract approaches in terms of ROUGE measurement. The result of the ablation study proves that the model has the ability to capture the original subject and the correct information and improve the factual accuracy of the summarization.

show abstract

Exploiting Discourse-Level Segmentation for Extractive Summarization

Cited by 23 publications

References 25 publications

Semantic Self-Segmentation for Abstractive Summarization of Long Documents in Low-Resource Regimes

Semantic Self-Segmentation for Abstractive Summarization of Long Documents in Low-Resource Regimes

Conditional Neural Generation using Sub-Aspect Functions for Extractive News Summarization

GATSum: Graph-Based Topic-Aware Abstract Text Summarization

Contact Info

Product

Resources

About