Publications
2025
A Framework for Supporting the Iterative Design of CBR Applications
Guillermo Jimenez-Diaz, Mirko Lenz, Lukas Malburg, Belén Díaz-Agudo, and Ralph Bergmann
Case-Based Reasoning Research and Development, ICCBR 2025, Lecture Notes in Computer Science, Vol. 15662, pp. 252-266, Springer Nature Switzerland, Cham
Abstract
Iterative design is a well-known methodology that involves prototyping, testing, analyzing, and refining products or processes. Rapid prototyping and evaluation are crucial, enabling designers to quickly identify and resolve issues and iteratively improve the design. However, effective iterative design relies on tools that accelerate both prototype creation and visual evaluation. This paper aims to address this challenge by presenting a framework specifically designed to enhance the iterative design process for CBR applications. In this context, we emphasize the manual and knowledge-intensive task of defining similarity measures. Furthermore, we introduce a proof-of-concept implementation of this framework based on the CBRkit toolkit and the SimViz visualization tool. We studied the capabilities to support the iterative design of CBR applications through a case study in the prototypical cars domain.
ArgueMapper Assistant: Interactive Argument Mining Using Generative Language Models
Mirko Lenz and Ralph Bergmann
Artificial Intelligence XLI, SGAI 2024, Vol. 15446, pp. 189-203, Springer Nature Switzerland, Cham
Abstract
Structured arguments are a valuable resource for analyzing and understanding complex topics. However, manual annotation is time-consuming and often not feasible for large datasets, and automated approaches are less accurate. To address this issue, we propose an interactive argument mining system that takes advantage of generative language models to support humans in the creation of argument graphs. We present the open source ArgueMapper Assistant featuring two prompting strategies and evaluate it on a real-world news dataset. The resulting corpus containing 88 argument graphs is publicly available as well. With generative models, the annotation time is reduced by about 20% while the number of errors is slightly increased (mostly due to missing argumentative units and wrong relation types). A survey provides insights into the usefulness and reliability of the assistant features and shows that participants prefer to use the assistant in the future.
Case-Based Reasoning Meets Large Language Models: A Research Manifesto For Open Challenges and Research Directions
Kerstin Bach, Ralph Bergmann, Florian Brand, Marta Caro-Martínez, Viktor Eisenstadt, Michael W. Floyd, Lasal Jayawardena, David Leake, Mirko Lenz, Lukas Malburg, David H. Ménager, Mirjam Minor, Brian Schack, Ian Watson, Kaitlynne Wilkerson, and Nirmalie Wiratunga
Abstract
In recent years, the surge of Generative Artificial Intelligence (GenAI ), particularly Large Language Models (LLMs), has led to a significant increase in the use of hybrid systems, which combine the strengths of different Artificial Intelligence (AI) paradigms to achieve better performance and efficiency. Although LLMs demonstrate remarkable effectiveness across numerous tasks due to their flexibility and general knowledge, they often face challenges related to accuracy, explainability, and their limited memory. Case-Based Reasoning (CBR), on the other hand, excels by recalling past experiences and using them to solve new problems, making it particularly well suited for tasks that require contextual understanding and decision-making. However, CBR systems suffer from issues such as the acquisition of various kinds of knowledge and the application of methods during the 4R cycle. In this paper, we identify several challenges plaguing LLMs and CBR systems and propose opportunities to combine the strengths of both methodologies to address these challenges. In addition, we outline future research directions for the community to explore.
EXAR: A Unified Experience-Grounded Agentic Reasoning Architecture
Ralph Bergmann, Florian Brand, Mirko Lenz, and Lukas Malburg
Case-Based Reasoning Research and Development, ICCBR 2025, Lecture Notes in Computer Science, Vol. 15662, pp. 3-17, Springer Nature Switzerland, Cham
Abstract
Current AI reasoning often relies on static pipelines (like the 4R cycle from Case-Based Reasoning (CBR) or standard Retrieval-Augmented Generation (RAG)) that limit adaptability. We argue it is time for a shift towards dynamic, experience-grounded agentic reasoning. This paper proposes EXAR, a new unified, experience-grounded architecture, conceptualizing reasoning not as a fixed sequence, but as a collaborative process orchestrated among specialized agents. EXAR integrates data and knowledge sources into a persistent Long-Term Memory utilized by diverse reasoning agents, which coordinate themselves via a Short-Term Memory. Governed by an Orchestrator and Meta Learner, EXAR enables flexible, context-aware reasoning strategies that adapt and improve over time, offering a blueprint for next-generation AI.
LLsiM: Large Language Models for Similarity Assessment in Case-Based Reasoning
Mirko Lenz, Maximilian Hoffmann, and Ralph Bergmann
Case-Based Reasoning Research and Development, ICCBR 2025, Lecture Notes in Computer Science, Vol. 15662, pp. 126-141, Springer Nature Switzerland, Cham
Abstract
In Case-Based Reasoning (CBR), past experience is used to solve new problems. Determining the most relevant cases is a crucial aspect of this process and is typically based on one or multiple manually-defined similarity measures, requiring deep domain knowledge. To overcome the knowledge-acquisition bottleneck, we propose the use of Large Language Models (LLMs) to automatically assess similarities between cases. We present three distinct approaches where the model is used for different tasks: (i) to predict similarity scores, (ii) to assess pairwise preferences, and (iii) to automatically configure similarity measures. Our conceptual work is accompanied by an open-source Python implementation that we use to evaluate the approaches on three different domains by comparing them to manually crafted similarity measures. Our results show that directly using LLM-based scores does not align well with the baseline rankings, but letting the LLM automatically configure the measures yields rankings that closely resemble the expert-defined ones.
2024
ArgServices: A Microservice-Based Architecture for Argumentation Machines
Mirko Lenz, Lorik Dumani, Ralf Schenkel, and Ralph Bergmann
Robust Argumentation Machines, RATIO 2024, Lecture Notes in Computer Science, Vol. 14638, pp. 352-369, Springer Nature Switzerland, Cham
Abstract
Argumentation is ubiquitous, and the development of argumentation machines could greatly assist humans in managing and navigating argumentation. However, the development of such systems is hindered by the lack of common standards and suitable tools, leading to ad-hoc solutions with little reuse value. Towards a more unified approach, we present an extensible microservice-based architecture for argumentation machines. Being built on the established gRPC framework, it provides strongly typed interfaces for the following services: (i) Argument Mining, (ii) Case-Based Reasoning on Arguments, (iii) Argument Retrieval and Ranking, and (iv) Quality Assessment of Arguments. Our system is designed to be extensible, allowing for easy integration of new tasks. We demonstrate the feasibility of our architecture via a proof-of-concept implementation and provide additional supplementary resources, such as a REST API gateway. Our contributions are publicly available on GitHub under the permissive MIT license.
CBRkit: An Intuitive Case-Based Reasoning Toolkit for Python
Mirko Lenz, Lukas Malburg, and Ralph Bergmann
Case-Based Reasoning Research and Development, ICCBR 2024, Lecture Notes in Computer Science, Vol. 14775, pp. 289-304, Springer Nature Switzerland, Cham
Best Student Paper Award at ICCBR 2024
Abstract
Developing Case-Based Reasoning (CBR) applications is a complex and demanding task that requires a lot of experience and a deep understanding of users. Additionally, current CBR frameworks are not as usable as Machine Learning (ML) frameworks that can be deployed with only a few lines of code. To address these problems and allow users to easily build hybrid Artificial Intelligence (AI) systems by combining CBR with techniques such as ML, we present the CBRkit library in this paper. CBRkit is a Python-based framework that provides generic and easily extensible functions to simplify the creation of CBR applications with advanced similarity measures and case representations. The framework is available from GitHub and PyPI under the permissive MIT license. An initial user study indicates that it is easily possible even for non-CBR experts and users who only have limited Python programming skills to develop their own customized CBR application.
PolArg: Unsupervised Polarity Prediction of Arguments in Real-Time Online Conversations
Mirko Lenz and Ralph Bergmann
Robust Argumentation Machines, RATIO 2024, Lecture Notes in Computer Science, Vol. 14638, pp. 108-126, Springer Nature Switzerland, Cham
Abstract
The increasing usage of social networks has led to a growing number of discussions on the Internet that are a valuable source of argumentation that occurs in real time. Such conversations are often made up of a large number of participants and are characterized by a fast pace. Platforms like X/Twitter and Hacker News (HN) allow users to respond to other users’ posts, leading to a tree-like structure. Previous work focused on training supervised models on datasets obtained from debate portals like Kialo where authors provide polarity labels (i.e., support/attack) together with their posts. Such classifiers may yield suboptimal predictions for the noisier posts from X or HN, so we propose unsupervised prompting strategies for large language models instead. Our experimental evaluation found this approach to be more effective for X conversations than a model fine-tuned on Kialo debates, but less effective for HN posts (which are more technical and less argumentative). Finally, we provide an open-source application for converting discussions on these platforms into argument graphs.
2023
Case-Based Adaptation of Argument Graphs with WordNet and Large Language Models
Mirko Lenz and Ralph Bergmann
Case-Based Reasoning Research and Development, ICCBR 2023, Lecture Notes in Computer Science, Vol. 14141, pp. 263-278, Springer Nature Switzerland, Cham
Best Student Paper Award at ICCBR 2023
Abstract
Finding information online is hard, even more so once you get into the domain of argumentation. There have been developments around the specialized argumentation machines that incorporate structural features of arguments, but all current approaches share one pitfall: They operate on a corpora of limited sizes. Consequently, it may happen that a user searches for a rather general term like cost increases, but the machine is only able to serve arguments concerned with rent increases. We aim to bridge this gap by introducing approaches to generalize/specialize a found argument using a combination of WordNet and Large Language Models. The techniques are evaluated on a new benchmark dataset with diverse queries using our fully featured implementation. Both the dataset and the code are publicly available on GitHub.
2022
Comparing Unsupervised Algorithms to Construct Argument Graphs
Mirko Lenz, Lorik Dumani, and Premtim Sahitaj
Joint Proceedings of Workshops, Tutorials and Doctoral Consortium Co-Located with the 45th German Conference on Artificial Intelligence, TMG 2022, CEUR Workshop Proceedings, Vol. 3457, CEUR, Virtual Event, Trier
Abstract
Computational argumentation has gained considerable attention in recent years. Various areas have been addressed, such as extracting arguments from natural language texts into a structured form in order to store them in an argument base, determining stances for arguments with respect to topics, determination of inferences from statements, and much more. After so much progress has been made in the isolated tasks, in this paper we address the next level and aim to advance the automatic generation of argument graphs. To this end, we investigate various unsupervised methods for constructing the graphs and measure the performance with different metrics on three different datasets. Our implementation is publicly available on GitHub under the permissive MIT license.
User-Centric Argument Mining with ArgueMapper and Arguebuf
Mirko Lenz and Ralph Bergmann
Computational Models of Argument, International Conference on Computational Models of Argument, Frontiers in Artificial Intelligence and Applications, Vol. 353, pp. 367-368, IOS Press, Cardiff, Wales
Abstract
Existing tools to create argument graphs are tailored for experts in the domain of argumentation. By taking into account the needs of experts, laymen, and developers, we propose ArgueMapper as a novel argument diagramming tool and Arguebuf as its underlying format. ArgueMapper is the first of its kind to be optimized for mobile devices and provide a discoverable interface suitable for novice users. Arguebuf provides native implementations for all major programming languages via a code generation approach. To complement Arguebuf, we provide a supercharged Python implementation that enables advanced analysis. All of our contributions support AIF and are publicly available on GitHub under the MIT license.
Workshop on Text Mining and Generation (TMG): Preface
Mirko Lenz, Lorik Dumani, Alexander Bondarenko, and Shahbaz Syed
Joint Proceedings of Workshops, Tutorials and Doctoral Consortium Co-Located with the 45th German Conference on Artificial Intelligence, TMG 2022, CEUR Workshop Proceedings, Vol. 3457, CEUR, Virtual Event, Trier
Abstract
This paper is a report on the first Text Mining and Generation Workshop (TMG), which was a one-day virtual event hosted at the German Conference on Artificial Intelligence (KI 2022) in Trier, Germany. In addition to four accepted original papers, there were three invited talks by speakers who presented their works already published at high-ranked conferences as well as one keynote by a pioneer in the two research fields relevant to the workshop.
2021
The ReCAP Corpus: A Corpus of Complex Argument Graphs on German Education Politics
Lorik Dumani, Manuel Biertz, Alex Witry, Anna-Katharina Ludwig, Mirko Lenz, Stefan Ollinger, Ralph Bergmann, and Ralf Schenkel
IEEE Proceedings of the 15th International Conference on Semantic Computing (ICSC), IEEE 15th International Conference on Semantic Computing (ICSC), pp. 248-255, IEEE, Laguna Hills, CA, USA
Abstract
The automatic extraction of arguments from natural language texts is a highly researched area and more important than ever today, as it is nearly impossible to manually capture all arguments on a controversial topic in a reasonable amount of time. For testing different algorithms such as the retrieval of the best arguments, which are still in their infancy, gold standards must exist. An argument consists of a claim or standpoint that is supported or opposed by at least one premise. The generic term for a claim or premise is Argumentative Discourse Unit (ADU). The relationships between ADUs can be specified by argument schemes and can lead to large graphs. This paper presents a corpus of 100 argument graphs with about 2,500 ADUs in German, which is unique in its size and the utilisation of argument schemes. The corpus is built from natural language texts like party press releases and parliamentary motions on education policies in the German federal states. Each high-quality text is presented by an argument graph and created by the use of a modified version of the annotation tool OVA. The final argument graphs resulted by merging two previously independently annotated graphs based on detailed discussions.
2020
The ReCAP Project
Ralph Bergmann, Manuel Biertz, Lorik Dumani, Mirko Lenz, Anna-Katharina Ludwig, Patrick J. Neumann, Stefan Ollinger, Premtim Sahitaj, Ralf Schenkel, and Alex Witry
Datenbank-Spektrum, Vol. 20, pp. 93-98
Abstract
Argumentation Machines search for arguments in natural language from information sources on the Web and reason with them on the knowledge level to actively support the deliberation and synthesis of arguments for a particular user query. The recap project is part of the Priority Program ratio and aims at novel contributions to and confluence of methods from information retrieval, knowledge representation, as well as case-based reasoning for the development of future argumentation machines. In this paper we summarise recent research results from the project. In particular, a new German corpus of 100 semantically annotated argument graphs from the domain of education politics has been created and is made available to the argumentation research community. Further, we discuss a comprehensive investigation in finding arguments and argument graphs. We introduce a probabilistic ranking framework for argument retrieval, i.e. for finding good premises for a designated claim. For finding argument graphs, we developed methods for case-based argument retrieval considering the graph structure of an argument together with textual and ontology-based similarity measures applied to claims, premises, and argument schemes.
Towards an Argument Mining Pipeline Transforming Texts to Argument Graphs
Mirko Lenz, Premtim Sahitaj, Sean Kallenberg, Christopher Coors, Lorik Dumani, Ralf Schenkel, and Ralph Bergmann
Computational Models of Argument, International Conference on Computational Models of Argument, Frontiers in Artificial Intelligence and Applications, Vol. 326, pp. 263-270, IOS Press, Virtual Event
Abstract
This paper tackles the automated extraction of components of argumentative information and their relations from natural language text. Moreover, we address a current lack of systems to provide a complete argumentative structure from arbitrary natural language text for general usage. We present an argument mining pipeline as a universally applicable approach for transforming German and English language texts to graph-based argument representations. We also introduce new methods for evaluating the performance based on existing benchmark argument structures. Our results show that the generated argument graphs can be beneficial to detect new connections between different statements of an argumentative text.
2019
Semantic Textual Similarity Measures for Case-Based Retrieval of Argument Graphs
Mirko Lenz, Stefan Ollinger, Premtim Sahitaj, and Ralph Bergmann
Case-Based Reasoning Research and Development, ICCBR 2019, Lecture Notes in Computer Science, Vol. 11680, pp. 219-234, Springer International Publishing, Otzenhausen, Germany
Abstract
Argumentation is an important sub-field of Artificial Intelligence, which involves computational methods for reasoning and decision making based on argumentative structures. This paper contributes to case-based reasoning with argument graphs in the standardized Argument Interchange Format by improving the similarity-based retrieval phase. We explore a large range of novel approaches for semantic textual similarity measures (both supervised and unsupervised) and use them in the context of a graph-based similarity measure for argument graphs. In addition, the use of an ontology-based semantic similarity measure for argumentation schemes is investigated. With a range of experiments we demonstrate the strengths and weaknesses of the various methods and show that our methods can improve over our previous work. Our code is publicly available on GitHub.
Similarity Measures for Case-Based Retrieval of Natural Language Argument Graphs in Argumentation Machines
Ralph Bergmann, Mirko Lenz, Stefan Ollinger, and Maximilian Pfister
Proceedings of the Thirty-Second International Florida Artificial Intelligence Research Society Conference, Thirty-Second International Florida Artificial Intelligence Research Society Conference, pp. 329-334, AAAI Press, Sarasota, Florida, USA
Abstract
In the field of argumentation, the vision of robust argumentation machines is investigated. They explore natural language arguments from information sources on the web and reason with them on the knowledge level to actively support the deliberation and synthesis of arguments for a particular user query. We aim at combining methods from case-based reasoning (CBR), information retrieval, and computational argumentation to contribute to the foundations of argumentation machines. In this paper, we focus on the retrieval phase of a CBR approach for an argumentation machine and propose similarity measures for arguments represented as argument graphs. We evaluate the similarity measures on a corpus of annotated micro texts and demonstrate the benefit of semantic similarity measures and the relevance of structural aspects.