Institute for Web Science and Technologies · Universität Koblenz - Landau
Institute WeST


Generating Counterfactual Images for Visual Question Answering by Editing Question-Critical Objects

17.06.21. While Visual Question Answering (VQA) systems improved significantly in recent years, they still tend to produce errors that are hard to reconstruct for human users. The lack of interpretability in black-box VQA models raises the necessity for discriminative explanations alongside the models’ outputs. This thesis aims at introducing a method to generate counterfactual images for an arbitrary VQA model. Given a question-image pair, the counterfactual generator should mask the question-critical objects in the image and then predict a minimal number of edits to the image such that the VQA model outputs a different answer. Thereby, the new image should contain semantically meaningful changes, be visually realistic, and remain unchanged in question-answer-irrelevant regions (e.g., the background). To the best of my knowledge, this is the first counterfactual image generator applied to VQA systems that does not apply edits to individual pixels but rather to a spatial mask without requiring additional manual annotations. [mehr lesen...]

Towards Explainable Creativity: Tackling the Remote Association Test with Knowledge Graphs

20.05.21. Creative problem solving is one of the topics that interest both Cognitive Scientists and researchers in Artificial Intelligence. One aim of cognitive scientists is to build various A.I. systems that can solve creative problems and answer questions like 'How the human mind works while solving a creative task?' For Artificial Intelligence, creative problem-solving systems can help modeling agents solve complex problems with novel ideas. Various frameworks can perform computational creativity tests like the Remote Association Test (RAT) to measure AI systems' cognitive and problem-solving abilities. However, these frameworks cannot propose explanations for these solutions. In this talk, I propose a topic for my master's thesis: Building an AI system that can solve creativity tests and propose an explanation for these solutions. [mehr lesen...]

Using Random Walk to Measure Inconsistency in Business Process Models

06.05.21. Business process models are used for the design and execution of workflows in organizations. These process models might be subject to various inconsistencies, for example due to violations of workflow norms. This thesis aims at providing modellers with means to firstly pinpoint errorneous elements, and secondly to quantify the risk emerging from inconsistent modelling. To this end, we formalize the execution semantics of business process models with the help of Petri nets and explore the state space of these nets in a probabilistic fashion. [mehr lesen...]

Approximate Inference for Assumption-based Argumentation in AI

15.04.21. This thesis focuses on approximate inference in assumption-based argumentation frameworks. Argumentation provides a significant idea in the computerization of theoretical and practical reasoning in AI. And it has a close connection with AI, engaging in arguments to perform scientific reasoning. The fundamental approach in this field is abstract argumentation frameworks developed by Dung. Assumption-based argumentation can be regarded as an instance of abstract argumentation with structured arguments. When facing a large scale of data, a challenge of reasoning in assumption-based argumentation is how to construct arguments and resolve at- tacks over a given claim with minimal cost of computation and acceptable accuracy at the same time. This thesis proposes and investigates approximate methods that randomly select and construct samples of frameworks based on graphical dispute derivations to solve this problem. The presented approach aims to improve reason- ing performance and get an acceptable trade-off between computational time and accuracy. The evaluation shows that for reasoning in assumption-based argumenta- tion, in general, the running time is reduced with the cost of slightly low accuracy by randomly sampling and constructing inference rules for potential arguments over a query. [mehr lesen...]

Measuring Disagreement with Interpolants

25.02.21. A disagreement measure is a function that quantitatively assesses the conflict between knowledge bases in knowledge merging scenarios. Using the notion of Craig interpolation we define a series of disagreement measures and analyse their compliance with properties proposed in previous work by Potyka. We study basic complexity-theoretic questions in that scenario and discuss the suitability of our approaches. [mehr lesen...]

Ältere Einträge