Aktuelle Lehrveranstaltungen

We usually teach the following courses:

Bachelor:

In the Bachelor, we lay the foundations for an understanding of tasks and methods for machine language understanding mainly with two lectures:

  • Information Retrieval and Text Mining: here we discuss how to build search engines, develop efficient data structures to answer search queries, sort documents by similarity, and categorize texts into predefined classes. Machine learning and deep learning techniques are introduced. Furthermore, this lecture serves as a first introduction to methods of natural language processing. The text unit that is the focus here is the document as a whole.
  • Natural Language Understanding (Algorithmic Language Understanding): Here we go deeper into the text and discuss methods to understand it more accurately, automatically. We start with methods that measure the meaning of individual words (vector representations, embeddings) and then extend this to analyzing phrases, including with (large) language models. We then discuss common tasks in language understanding, such as entity recognition, relation recognition, emotion/sentiment analysis, argument analysis, but also tasks as solved by large instruction-based trained language models.

Master:

In the Master's programme, we deepen and specialize the fundamentals acquired in the Bachelor's programme. We currently offer the following lectures on a recurring basis:

  • Probabilistic Graphical Models for Natural Language Processing: Many problems of language comprehension are structure learning problems that can be simplified and modelled by factorizations of probability distributions. In this course we will review the basics of probability theory, then extend it with methods of directed graphical models and undirected models. We learn methods of efficient inference and parameter and structure learning and apply them to problems of language processing.
  • Societal Impacts of Language Technology: Language processing systems have the potential to have a major impact on our society. It is therefore all the more important to question where this role must be viewed critically. What happens if the training data of a system mainly represents a subset of the population? What does the system then do to other users? Can such automated systems be trusted? These questions will be discussed in this course.
  • Emotion Analysis: Emotion analysis is the task of reconstructing the emotions of an author, for example of a tweet, or estimating the effect of a news text on readers. To build such systems, in this lecture we first develop an understanding of emotion theories in psychology, we then discuss how to create data and models and use the knowledge available in psychology. Furthermore, we present different applications.

In addition, we offer special seminars:

  • Explainable AI for NLP Models: In this seminar we will discuss XAI procedures. These are generally applicable, but pose some special challenges in the field of NLP.
  • Argument Mining: Where does a text state an argument? Are there conditions under which it is valid? What is the quality of the argument? These questions can be answered automatically with the help of a computer, at least to some extent. In this seminar we will discuss argument theories and modeling.
  • Large Language Models for Natural Language Understanding: We will discuss large language models, how to create and optimize them and how to query them. These prompts can be created automatically or manually. In the automatic case, we are talking about prompt optimization. These methods have partly replaced classical machine and deep learning to solve speech understanding problems. We discuss the latest publications on this topic here.
  • Multimodal text analysis: For a long time, the analysis of social media or news texts has focused on text alone. Sometimes, however, it is necessary to analyze the images together with the text. We will discuss methods that make this possible here, with a focus on the text analysis perspective.

In the winter semester 2024/2025 we offer:

Master Seminars:

  • Large Language Models for Natural Language Understanding
  • Explainable AI for NLP Models
  • Multimodal Text Analysis

Master Lectures:

  • Societal Impacts of Language Technology
  • Probabilistic Graphical Models for Natural Language Processing

Bachelor lectures:

  • Natural Language Understanding (Algorithmic Language Understanding)

Regarding the registration for seminars: We offer two seminar modules, such that you can take two seminars from our research group. If you want more, we offer, in addition, a 6 ECTS module that you can fill with two smaller courses, including two seminars.

Below we show the information automatically extracted from Univis. However, these are incorrect at the time this text was created. They will probably be corrected relatively soon.

 

Vorlesungen

NLPROC: Natural Language Understanding
2 SWS
Roman Klinger
NLPROC: Probabilistic Graphical Models for Natural Language Processing
4 SWS
Sean Papay
NLPROC: Societal Impacts of Language Technology
4 SWS
Sabine Weber

Übungen

NLPROC: Natural Language Understanding
2 SWS
Roman Klinger

Seminare

NLPROC: Argument Mining in Natural Language Processing
2 SWS
Lynn Greschner , Roman Klinger
NLPROC: Explainable AI for NLP Models
2 SWS
Sabine Weber , Sean Papay
NLPROC: Kolloquium begleitend zu Abschlussarbeit
2 SWS
Roman Klinger
NLPROC: Large Language Models for Natural Language Understanding
2 SWS
Roman Klinger
NLPROC: Multimodal Text Analysis
2 SWS
Christopher Bagdon , Roman Klinger