ACCEPT: Perspectivized Argument Knowledge Graphs for Deliberation Support
A joint project between Prof. Anette Frank (ICL, Heidelberg University)
and Prof. Philipp Cimiano
(University of Bielefeld) in the 2nd phase of the DFG priority program RATIO:
Robust Argumentation Machines
In ACCEPT we aim to create systems that are able to perform deep understanding of debated issues and analysis of
in order to find optimal and widely accepted solutions on debated issues.
The project aims to push the limits beyond prior work in computational argumentation, by designing systems that are able to
We aim to do this
- by creating argumentation systems that leverage knowledge resources to perform knowledge-based reasoning, using both sub-symbolic modeling using neural models and symbolic modeling using knowledge graphs, and
- by representing, contextualizing and enriching arguments in a multi-factorial argument knowledge graph that includes stakeholder perspectives and their interests, values and goals.
Based on the argument knowledge graph we will develop methods that analyze debated issues from multiple perspectives and learn to reason towards solutions, in order to support users with interpretable deliberation support.
We build on own research conducted in the ExpLAIN project in the first project phase of RATIO. Plesase consult our webpage.
NEWS
Sep 23: Following up to the RATIO Hackathon event at Bielefeld University in April, Moritz Plenz, Raphael Buchmüller and Alexander Bondarenko, all part of the SPP RATIO project, participated in the Touché Argument and Causal Retrieval Task at CLEF 2023 with their work on Argument Quality Prediction for Ranking Documents, which they presented in Thessaloniki in September.
July 13-14: ACCEPT will present new work by Philipp Heinisch, Moritz Plenz, Anette Frank and Philipp Cimiano: An Ensemble-based Approach to Multilingual Framing Detection at SemEval-2023, co-located with ACL 2023 in Toronto.
The problem: Predicting the generic frame set from a lengthy (multilingual) news article poses a significant challenge.
Our solution: By employing an ensemble of a Graph Neural Networks (GNNs) that process background-knowledge enriched graphs, along with simple neural networks utilizing static word embeddings (which exhibit surprisingly strong performance), we can achieve consistent and reliable results.
July 10-12: ACCEPT will present Similarity-weighted Construction of Contextualized Commonsense Knowledge Graphs for Knowledge-intense Argumentation Tasks, by Moritz Plenz, Juri Opitz, Philipp Heinisch, Philipp Cimiano and Anette Frank as a long paper at ACL 2023 in Toronto.
Abstract:We present a method to construct Contextualized Commonsense Knowledge Graphs (CCKGs). The method goes beyond context-insensitive knowledge extraction heuristics by computing semantic similarity between KG triplets and textual arguments to extract contextualized knowledge paths that connect a conclusion to its premise, while maximizing similarity to the argument.
June 30: PhD defense!!
We're celebrating the successful PhD defense of Juri Opitz,
for his thesis on Metrics of Graph-Based Meaning Representations with Applications from
Parsing Evaluation to Explainable NLG Evaluation and Semantic Search !
Congratulations, Juri, to this very rich and fine work! It has greatly contributed to our research in the ACCEPT project!
Thanks for your great work and collaboration, and best of luck for the next steps in your career!!
Apr 24-28: ACCEPT participates in the RATIO Hackathon at Bielefeld University, participating in two projects:
Argument Quality Prediction for Ranking Documents in the Touché Argument and Causal Retrieval Task at CLEF 2023, and in the Frame Prediction Task.
Jan 12-13: Gathering in Bielefeld for our bi-annual project-internal ACCEPT workshop, to discuss and plan our next project milestones.
Nice to gather 'live' around a table, entertained with games and enjoying a lovely project dinner!
We presented two new publications in the Argument Mining Workshop 2022:
- Data Augmentation for Improving the Prediction of Validity and Novelty of Argumentative Conclusions by Philipp Heinisch et al., which was selected as
Runner-up for Best Paper Award! Congratulations! and - Overview of the 2022 Validity and Novelty Prediction Shared Task, with Proceedings of the ValNov Shared Task Contribution Papers.
Oct 4-6: ACCEPT participates in the RATIO 2022 Symposium in Bielefeld, which features Workshops on 1) Perspectives of Argumentation, 2) Inference & Summarization in Argumentation and 3) Domain-Specific Argumentation.
The Validity and Novelty Prediction Shared Task results are published. Six teams participated with overall 15 system submissions.
ACCEPT organizes a Shared Task on Predicting the validity and novelty of arguments in the ArgMining Workshop at COLING 2022. Task description and data are ready for participation!
New paper by Philipp Heinisch and members of ACCEPT
on Strategies for Framing Argumentative Conclusion Generation,
to be presented at the 15th International Natural Language Generation Conference (INLG 2022).
Congratulations to Juri Opitz and all team members for earning the
Best Paper
Award in the Eighth ArgMining
Workshop@EMNLP, for Explainable Unsupervised Argument
Similarity Rating with Abstract Meaning Representation and Conclusion Generation !
Anette Frank will give a panel talk in the Eighth ArgMining Workshop @EMNLP.
Philipp Cimiano will give an invited talk at ArgKG @ AKBC 2021.
Anette Frank will give an invited talk at CLAR 2021, the 4th International Conference on Logic and Argumentation, Hangzhou, China.
Two papers accepted from members of ACCEPT
- Explainable Unsupervised Argument Similarity Rating with Abstract Meaning Representation and Conclusion Generation, by Juri Opitz, Philipp Heinisch, Philipp Wiesenbach, Philipp Cimiano and Anette Frank, in the ArgMining Workshop @EMNLP.
- Key Point Analysis via Contrastive Learning and Extractive Argument Summarization, by Milad Alshomary, Timon Gurcke, Shahbaz Syed, Philipp Heinisch, Maximilian Spliethöver, Philipp Cimiano, Martin Potthast, and Henning Wachsmuth in the ArgMining Workshop @EMNLP.
PEOPLE
PIs
Prof. Anette Frank
Head of the NLP Group Heidelberg University| Email (click to copy) | |
| Semantic NLP for advanced & situated Natural Language Understanding |
Prof. Philipp Cimiano
Head of Semantic Computing Group University of Bielefeld| Email (click to copy) | |
| Knowledge Representation and Natural Language Processing |
Doctoral Researchers
Juri Opitz
MA Computational Linguistics PhD Student Computational Linguistics| Email (click to copy) | Heidelberg University |
Moritz Plenz
MSc Physics PhD Student Computational Linguistics| Email (click to copy) | Heidelberg University |
SELECTED PUBLICATIONS
Data Augmentation for Improving the Prediction of Validity and Novelty of Argumentative Conclusions
Heinisch, P., Plenz, M., Opitz, J., Frank, A., Cimiano, P. (2022)
Proceedings of the 9th Workshop on Argument Mining.
Overview of the 2022 Validity and Novelty Prediction Shared Task
Heinisch, P., Frank, A., Opitz, J., Plenz, M., Cimiano, P. (2022)
Proceedings of the 9th Workshop on Argument Mining.
Strategies for Framing Argumentative Conclusion Generation
Heinisch, P., Frank, A., Opitz, J., Cimiano, P. (2022)
Proceedings of the 15th International Natural Language Generation Conference
Weisfeiler-Leman in the BAMBOO: Novel AMR
Graph Metrics and a Benchmark for AMR Graph Similarity
Opitz, J., Daza, A., Frank, A. (2021)
Transactions of the Association for Computational Linguistics (TACL).
Explainable Unsupervised Argument Similarity Rating with Abstract Meaning Representation and Conclusion Generation
Opitz, J., Heinisch, P., Wiesenbach, P., Cimiano, P., Frank, A. (2021)
Proceedings of the Eighth ArgMining Workshop, Punta Cana, Dominican Republic.
