Key Point Analysis via Contrastive Learning and Extractive Argument Summarization

Alshomary, Milad; Gurcke, Timon; Syed, Shahbaz; Heinisch, Philipp; Spliethöver, Maximilian; Cimiano, Philipp; Potthast, Martin; Wachsmuth, Henning

doi:10.18653/v1/2021.argmining-1.19

Cited by 5 publications

(4 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use pair classification for our models in KPM experiments, which gives better results than contrastive learning. For KPM, we select 5 best results from Friedman et al (2021): SMatchToPR (Alshomary et al, 2021) and MatchTstm (Phan et al, 2021) use contrastive learning; NLP@UIT, Enigma (Kapadnis et al, 2021) andModernTalk (Reimer et al, 2021) use pair classification. For CESC, we select the results of both approaches mentioned in Cheng et al (2022).…”

Section: Resultsmentioning

confidence: 99%

Hi-ArG: Exploring the Integration of Hierarchical Argumentation Graphs in Language Pretraining

Liang,

Ye,

Han

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

The knowledge graph is a structure to store and represent knowledge, and recent studies have discussed its capability to assist language models for various applications. Some variations of knowledge graphs aim to record arguments and their relations for computational argumentation tasks. However, many must simplify semantic types to fit specific schemas, thus losing flexibility and expression ability. In this paper, we propose the Hierarchical Argumentation Graph (Hi-ArG), a new structure to organize arguments. We also introduce two approaches to exploit Hi-ArG, including a text-graph multi-modal model GreaseArG and a new pre-training framework augmented with graph information. Experiments on two argumentation tasks have shown that after further pre-training and fine-tuning, GreaseArG supersedes same-scale language models on these tasks, while incorporating graph information during further pre-training can also improve the performance of vanilla language models. Code for this paper is available at https://github.com/ljcleo/Hi-ArG.

show abstract

Section: Resultsmentioning

confidence: 99%

Hi-ArG: Exploring the Integration of Hierarchical Argumentation Graphs in Language Pretraining

Liang,

Ye,

Han

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…Our approach, based on topic modeling and enhanced by hyperparameter tuning, not only segments the data space but also approximates the frequency of statements associated with each key point, filling a gap in previous KPG methodologies (Bar-Haim et al, 2021 ). Our decision to adopt an abstractive summarization approach, as opposed to traditional extractive methods (Bar-Haim et al, 2020b , 2021 ; Alshomary et al, 2021 ), proved advantageous. It allowed for a broader semantic representation within key points, capturing different aspects of a subtopic more effectively.…”

Section: Discussionmentioning

confidence: 99%

“…Finally, following the ArgMining Workshop 2021, a corresponding training and evaluation data set (Friedman et al, 2021 ) was published, explicitly designed for the development of new KPA methods. Based on this data set, further extractive approaches of KPG have been developed, which solve the problem via a graph-based method (Alshomary et al, 2021 ) or via the selection of representative key point candidates with the help of a combination of the evaluation metric MoverScore (Zhao et al, 2019 ) and the maximal marginal relevance (MMR) based on word embeddings of the statements (Shirafuji et al, 2021 ).…”

Section: Introductionmentioning

confidence: 99%

Key point generation as an instrument for generating core statements of a political debate on Twitter

Ehnert,

Schröter

2024

Front. Artif. Intell.

View full text Add to dashboard Cite

Identifying key statements in large volumes of short, user-generated texts is essential for decision-makers to quickly grasp their key content. To address this need, this research introduces a novel abstractive key point generation (KPG) approach applicable to unlabeled text corpora, using an unsupervised approach, a feature not yet seen in existing abstractive KPG methods. The proposed method uniquely combines topic modeling for unsupervised data space segmentation with abstractive summarization techniques to efficiently generate semantically representative key points from text collections. This is further enhanced by hyperparameter tuning to optimize both the topic modeling and abstractive summarization processes. The hyperparameter tuning of the topic modeling aims at making the cluster assignment more deterministic as the probabilistic nature of the process would otherwise lead to high variability in the output. The abstractive summarization process is optimized using a Davies-Bouldin Index specifically adapted to this use case, so that the generated key points more accurately reflect the characteristic properties of this cluster. In addition, our research recommends an automated evaluation that provides a quantitative complement to the traditional qualitative analysis of KPG. This method regards KPG as a specialized form of Multidocument summarization (MDS) and employs both word-based and word-embedding-based metrics for evaluation. These criteria allow for a comprehensive and nuanced analysis of the KPG output. Demonstrated through application to a political debate on Twitter, the versatility of this approach extends to various domains, such as product review analysis and survey evaluation. This research not only paves the way for innovative development in abstractive KPG methods but also sets a benchmark for their evaluation.

show abstract

“…SMatchToPageRank (hereafter as SMatch-ToPR) (Alshomary et al, 2021) aims to learn an embedding space where matching argument-KP pairs are closer than non-matching pairs. This model has two inputs: (1) the argument; and (2) the concatenation of the KP and the topic.…”

Section: Matching Trackmentioning

confidence: 99%

Overview of the 2021 Key Point Analysis Shared Task

Friedman¹,

Dankin²,

Hou³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

Key Point Analysis via Contrastive Learning and Extractive Argument Summarization

Cited by 5 publications

References 15 publications

Hi-ArG: Exploring the Integration of Hierarchical Argumentation Graphs in Language Pretraining

Hi-ArG: Exploring the Integration of Hierarchical Argumentation Graphs in Language Pretraining

Key point generation as an instrument for generating core statements of a political debate on Twitter

Overview of the 2021 Key Point Analysis Shared Task

Contact Info

Product

Resources

About