Master's Thesis: Explainability of Transformer Models in Authorship Verification

Background/Motivation: Authorship verification (AV) is used in areas such as forensics, plagiarism detection, and fake news detection to identify the true author of a text. The goal of authorship verification (AV) is to classify whether two or more texts were written by the same author (Y) or not (N). As in most AI fields today, the most powerful models are usually based on transformer architectures. While these continuously achieve new state-of-the-art results, the application of explainability methods is mostly limited to somewhat older models or architectures. This limits the applicability of the latest methods in practice.

Objective: The objective of this work is to apply existing explainability methods to newer, more powerful models and, if necessary, adapt them accordingly. Furthermore, the corresponding methods should be extended, if necessary, to be understandable for non-experts as well. This can be achieved, for example, thru visualizations or the automated extraction of the most important input data.

Results: The work should illustrate which explainability approaches are suitable for transformer models or can be adapted for them. Additionally, the various explainability approaches should be compared and, if necessary, combined to create an explainability framework that can be applied in practice. As a result, the findings provide both a scientific contribution to the explainability of modern transformer models and a direct practical benefit in the application of the methods.

Be part of change

• Implementation of one or more current models for authorship verification.

• Research and implementation of a novel explainability method for transformer models in the context of authorship verification (AV).

• Detailed evaluation and comparison of the method with existing SOTA methods on standard datasets (e.g., PAN).

What you contribute

• Very good academic record with a maximum average grade of 2.2

• Very good programming skills in Python are strictly required and will be assessed through practical tasks during the selection process

• Knowledge in Machine Learning, including training, inference, and optimization of transformer architectures

• Knowledge in ML-Security is desirable

• Good Python skills, specifically with PyTorch

• Scientific interest and interest in current research projects

What we offer

• Independent work schedule management

• Insights into the intersection of academic research and industrial application

This position is also available on a part-time basis. We value and promote the diversity of our employees' skills and therefore welcome all applications – regardless of age, gender, nationality, ethnic and social origin, religion, ideology, disability, sexual orientation and identity. Severely disabled persons are given preference in the event of equal suitability. Our tasks are diverse and adaptable – for applicants with disabilities, we work together to find solutions that best promote their abilities.

With its focus on developing key technologies that are vital for the future and enabling the commercial utilization of this work by business and industry, Fraunhofer plays a central role in the innovation process. As a pioneer and catalyst for groundbreaking developments and scientific excellence, Fraunhofer helps shape society now and in the future.

Ready for a change? Then apply now and make a difference! Once we have received your online application, you will receive an automatic confirmation of receipt. We will then get back to you as soon as possible and let you know what happens next.

Fraunhofer Institute for Secure Information Technology SIT www.sit.fraunhofer.de

Requisition Number: 82684 Application Deadline: